Friday, February 09, 2007

A Strange, Long, and Rambling Post....

wherein I start out criticizing Dembski, in the end sort-of agree with him, and in the middle repeat many Cosmological ID arguments I have made before.

Over on UD, Bill Dembski is upset at some criticism leveled at his design inference. He opens with:

Here’s a critique of the mathematics of the design inference from an assistant professor of religious studies. The combination of ignorance and arrogance on the part of this individual is staggering.

It would be tempting to base an entire post on the amusing aspect of Dembski leveling a charge of arrogance, and fill it with analogies along the lines of "that's like Imelda Marcus complaining that Leona Helmsley has too many shoes" but—well, it'd be too easy.

Let's look at the post that has Dembski upset:

Youre probably referring to the pseudo-mathematical posturings of William Dembski. Dembski is a fraud whom nobody should take seriously. Here's why: Dembski's model of specified complexity assumes that when attempting to determine the likelihood of a given pattern coming about randomly, that you have the pattern in mind from the outset. In other words, that evolution is a teleological process. But evolution is NOT teleological. It is not more unlikely, from a mathematical perspective, that, say, an eye should develop from a process of natural selection than that some other arbitrary result should take place. It's only mathematically unlikely because you are separating this singular event (i.e., the one that took place), from the billions of other equally singular events that COULD have taken place, but didn't. Those events were equally unlikely. PROSPECTIVELY, any one of them could have occurred. It's only RETROSPECTIVELY that we look at the one that did and say it's unlikely. . . .

Scott Paeth, PhD.
Assistant Professor of Religious Studies
DePaul University

Imagine my disappointment when, in this case, Dembski is correct. This is, indeed, a breathtakingly awful criticism. At first I thought he was going to make a reasonable argument—perhaps pointing out that for anything real, such as a flagellum, Dembski's design filter, were it to be applied, which is never has (we're all waiting—apparently for hell to freeze over) must at some point beg the question—effectively it must assume the flagellum is irreducibly complex before it can prove it was designed.

Paeth doesn’t go there, instead he makes one of the most annoying arguments that anyone, anywhere makes against design. And he uses ALL CAPITALS when he wants to be sure we get his point.

This oft-misused argument should be called the: I know enough probability that I am aware that a royal flush is no less likely than any other specific five-card hand argument. Which is true, and is even an important argument in areas like statistical mechanics, but it often gets applied in an incorrect manner, namely: nothing macroscopic should ever be surprising because no matter how amazing it is, it is but one of any number of equally unlikely possible outcomes that, had they occurred, would look just as amazing. Therefore, only chauvinism causes us to feel privileged.

OK, I'm about to take an unplanned diversion into Cosmological ID land. If you are not interested, skip down to the paragraph that begins with: "Back to Paeth."

When applied to cosmological ID this is a bad argument for at least two reasons.
  1. It's wrong. In terms of the physical constants, perhaps it's true that the constants are the result of a random draw (superstring landscape). However, although, in that scenario, all draws are equally unlikely, very few draws are winning hands, meaning they lead to a habitable universe (for any kind of life.) Thus, we still have a right to be amazed at our royal flush—unless there is a vast sample space of universes in existence, then we agree that our amazement has no basis.

  2. (The more important point) It's irrelevant for the Cosmological ID argument, since the CID argument is not (I know, I repeat myself) a low-probability argument, but a high-sensitivity argument. To quote myself from a comment on another blog:
    The fine-tuning/cosmological ID argument is not based on improbability, even though it is often expressed that way. It is based on sensitivity. As many have correctly pointed out, there is no way to calculate the probability of various constants. The fact that many find striking (regardless of their ID position) is that if you change their values by small amounts, the universe goes to hell, so to speak.

    It's tempting, but not correct, to relate this to improbability. However, at least in my opinion, almost the opposite is true. The more unlikely the constants, the more they appear to come from a random draw, then the more indirect credibility (again, this is my opinion) for non-ID explanations, such as the superstring landscape.

    The best thing that could ever happen for cosmological ID is a fundamental theory that predicts the values of the constants. In that case, they would not be improbable but, on the contrary, have a probability of unity. That would mean, given life's sensitivity to their values, that habitability was built, in the form of those laws explaining the constants, into the very fabric of the universe. That would be an enormous bit of circumstantial evidence in favor of cosmological ID.

Somehow my meanderings took me a talk-origins article that misapplies the probability argument with almost unheard boldness. The writer, one Nathan Urban makes the same two blunders, and with such gusto! He wrote, in an aging Talk Origins Post of the Month (yikes!)

And the claim "either the Universe was designed specifically for us by a creator or there is a multitude of universes" is clearly a false dichotomy -- there are other possibilities, perhaps more plausible than either of those two.

Yes and no. There is certainly at least this third possibility: there is just a single universe and we are very lucky indeed. Apart from that, what possibilities exist besides design and multiple—either simultaneous or in sequence or both—universes? Urban claims there are others, perhaps more plausible, but like a lost proof of Fermat's last theorem, he must have written them only in the margins. I'd like to know. Susskind, who argues that if the landscape is wrong it will be hard to answer the IDers, would also, presumably, like to know.

Urban writes:

Sure, the parameters are "fine-tuned" to produce life, but who says that the parameters could have taken on any other values in the first place?? If you're going to say that it's "improbable" that such a universe could have arisen, you must presuppose that the universe could have evolved some other way, but we have no information whatsoever on how, if at all, that may have occurred. It could be a law of physics that the constants could only take on the values that they do, for all we know!

See! Both errors, almost using just one sentence! Amazing! He misses the boat, of course, because he ties the design to improbability. That may be the right thing to do in biology, but in cosmology it's not. Indeed, as I wrote above, a universe where the constants are determined is a net win for the design view, and a crushing blow to theories such as the superstring landscape. Susskind tacitly recognizes this and is pushing his theory even though the price of the superstring landscape is high, and includes: (a) changing the definition of science to include the untestable and (b) proclaiming the death of physics—i.e., Susskind concludes (since design cannot be right) that there will never be a fundamental theory that explains the constants. Searching for such a theory, according to Susskind, is a fool's errand not unlike religion. He may be more right than he realizes.

Urban also writes:

The same goes for the laws of physics themselves. Who says it's even possible for the universe to exist in dimensions other than four? It's very likely might not be; there are many mathematically unique things about four dimensions, and the same laws of physics simply might not exist at all with any other number -- they might not generalize to arbitrary dimensions. It would thus make no sense to say that the universe is "fine-tuned" to four dimensions, since it couldn't be any other way.

This is very muddled and not unrelated logic. He argues, if I parse correctly, that (a) we need a universe with four dimensions (correct) but (b) how can that be a sign of design because who says any other dimensioned universe is even possible? (instead of design, he attributes our fortune to the "mathematical uniqueness" of four dimensions.) Again he misses the boat, namely: if no other dimensionality is even possible, that is a net plus for design. Once again, string theory indirectly backs this up: it makes no claim as to why exactly four of its many dimensions kept on expanding—perhaps in some universes it's a different number. So, if the universe could have a different number of dimensions, that would be bad for design for no other reason than it would be good for string theory. Urban says the opposite: if the universe is fundamentally restricted to dimensions it hurts design. No, it actually hurts the multiverse theory.

He then goes on to discuss Smolin's cosmic evolution. This cannot be one of his alternatives he alluded to earlier (although I think he thinks it is) when complaining of false dichotomies, because cosmic evolution is certainly a multiple universe theory (in fact my personal favorite.) Urban writes:

Second, even if the parameters were fine-tuned, who says that the "fine-tuner" is intelligent? The universe could fine-tune itself. Self-organizing critical systems are capable of fine-tuning all by themselves, following only a simple set of physical laws -- thus making it likely that the parameters are "fine-tuned" the way we see them.

Lee Smolin is attempting to verify such a theory, which he calls "cosmological natural selection". (This is a real, falsifiable physical theory. Quantum gravity would be required to support some of its basic hypotheses, but is not required to support its predictions. So far it has passed the tests which have been applied to it, though that's by no means conclusive.) Cosmological natural selection makes predictions -- for example, it predicts that we should expect universes with stars to be highly probable.

Look at the main prediction that Urban relates: Smolin predicts that we should expect universes with stars to be highly probable.

That is a indeed a "prediction" of Smolin's theory, but in the same sense that string theory "predicts" gravity. Smolin built in the answer by an ansatz: black holes produce new universes with very similar physics (constants). Given this assumption, natural selection will be biased toward producing universes with many black holes (since they are good at passing along their genes.) So we can expect a drift, over time, to a cosmos with many stars, which are prepubescent black holes. Such universes are, as a lucky consequence, habitable. Arguing, as Urban does, that universes with many stars constitute a prediction of Smolin's theory is a shell game—his prediction is built into the theory from the onset. Smolin does make some rather vague (and to me unsatisfying predictions) that boil down to this: if we can prove our universe isn't the best or nearly the best of all possible universes for producing black holes, his theory, he claims, will have been falsified.

Urban then writes:

Of course the theories I've mentioned are still rather speculative, but they certainly show that an intelligent designer is not a logical necessity. (And they also don't require a "multitude of universes" or "multiverse".)

Well, I read the article three times and the only theory he mentions is Smolin's—and if that doesn't require a multitude of universes then, for the same reason, evolution should have needed just a couple generations to take us from singled celled organisms to humans.

At the end, just like most before him and most after him, Urban, explicitly invokes the dreaded "every draw is equal, every draw is beautiful" argument:

As a similar example, look at it this way: suppose hypothetically that the parameters of the universe were determined purely at random by some natural physical process (without intelligent design being involved), such as a quantum fluctuation or something. Further suppose that there are 10 such parameters, which can take on values between 1 and 6, with every permutation being equally likely. And finally suppose that the only configuration of parameters capable of giving rise to a universe with intelligent life is 3526525514, and that the universe happens to, by random, come up with that configuration. To us, those parameters are a meaningless and random sequence, no more and no less likely than any other. But to them, it's an extremely special, unique, and very improbably "fine-tuned" -- the odds are worse than 60 million to one! -- set of parameters. But it would be incorrect for them to conclude that their universe was intelligently designed, because in this hypothetical example, it wasn't! (And again, this does not require a "multiverse".) No matter what configuration actually occurs, you can always after the fact say that that configuration was "selected for" simply by virtue of it being so improbable and you being in it, when in fact it's no more improbable than any other!

Again, and I know I am now repeating the repititions, Urban is arguing that the improbability of our constants is what leads IDers to invoke design. Wrong. Now, it's perhaps mean-spirited to be so harsh with Urban, because many Cosmological IDers make the same mistake.

Perhaps an illustration will help to demonstrate my point.

Instead of Urban's randomly selected universe, let me postulate another made up scenario:
  1. In our make believe universe, it turns out that if some constant C differed by one part in 10100 it would not be suitable for any kind of life (no stars, no heavy elements)

  2. We have just discovered "the" fundamental law of physics, and it predicts C down to that necessary 100th point and beyond.

Now that would be a great day for Cosmological Intelligent Design, even though C, in that universe, is the very opposite of improbable. At that point, the debate could only be between those who interpreted the universe's built-into-the-fabric-of-spacetime habitability as design and those who saw it as blind luck. The "random draw" alternative, as provide by the superstring landscape, would no longer be an option.

Back to Paeth who, irrespective of misconceptions he may have about "specified complexity" is simply joining a long line of fellow travelers misapplying the "all hands are equally likely" argument. Dembski is right to call him on it.

But while we are over on UD, we should take a look at the comments for this post (looking being the only thing I'm allowed to do at UD.)

One commenter uses Paeth's argument as yet another example as to why scientists are not to be trusted—why it may even cause him to abandon his believe in an old universe and a local Noahic flood. Now, I ask, why would anything that Paeth writes, given he is a professor of Religious Studies, influence your views on the merits of a scientific theory (an old universe)? Why, it's akin to PZ fretting over what Scott Adams has to say about evolution.

Another commenter sarcastically points out Paeth's fish-out-of-water problem, that he (Paeth) is a religious studies professor, so how could he ever hope to educate a mathematician (Dembski) on mathematics? There is some merit to that position—training does matter—although you would think that on UD, where there are a slew of non-biologists (including Dembski) lecturing professional biologists about the field of evolutionary biology, that there would be a certain reluctance or at least humility involved when unleashing that particular criticism.

No comments:

Post a Comment