Monday, August 29, 2005

From the horse's mouth

I don't like intelligent design as an idea. Sure, things are designed, but the claim that this or that is so complex or whatever is a bit too facile, and it may just cross on over into blind stupidity. It's a surrender at a key point. Consider this letter sent by several IDC bigwigs to Science Magazine, the leading American scientific journal.

Alan I. Leshner (“Redefining Science,” July 8) says intelligent design isn’t science because scientific theories “explain what can be observed” and are “testable by repeatable observations and experimentation.” But particular design arguments meet this standard. Biologist Michael Behe, for instance, argues that design is detectable in the bacterial flagellum because the tiny motor needs all its parts to function—is irreducibly complex—a hallmark of designed systems. The argument rests on what we know about designed systems, and from our growing knowledge of the cellular world and its many mechanisms.

How to test and discredit Behe’s argument? Provide a continuously functional evolutionary pathway from simple ancestor to present motor. Darwinists like Kenneth Miller point to the hope of future discoveries, and to the type III secretory system as a machine possibly co-opted on the evolutionary path to the flagellum. The argument is riddled with problems, but it shows that Miller, at least, understands perfectly well that Behe’s argument is testable.
No, you band of putzes! Miller proposed a testable evolutionary hypothesis. IDC proposed an untestable hypothesis: Some intelligent designer did this.

If Miller's theory is falsified, it doesn't mean that no evolutionary pathway exists. If it isn't falsified, it doesn't make it true. We will never be 100% certain about how the flagellum came to be, just as we will never be 100% certain that any scientific hypothesis is absolutely true. The best we can do is say that it is consistent with the data.

For centuries, all the data were consistent with Newton's laws of motion, until Einstein came along, proposed a new model and a test. If he was right, certain observations would be made. If he was wrong and Newton was right, other observations would be made. If both were wrong, entirely different results would be made (with some small probability that the other hypotheses might give results identical to those predicted by either competing model). Einstein's predictions proved to be accurate. Newton's model of the universe was incomplete – wrong, if you will.

Evolution predicts that we will find a series of molecular structures similar to the flagellum but with reduced numbers of parts. It predicts that, with fewer proteins and slight modifications to the remaining proteins, the structure will do something that helps the bacteria, or at least doesn't do it any harm. We expect that some of the proteins will have homologues (proteins with similar but distinct structures) which are membrane bound.

In fact, many of those predictions are accurate.

IDC, by contrast, offers this as a "hypothesis": No evolutionary pathway which results in a flagellum is possible. In order to test that hypothesis, you'd have to evaluate every evolutionary hypothesis, a search space which is infinite, though it may be bounded. Indeed, the claim "some evolutionary pathway exists" is verifiable, but not falsifiable, hence, not within the realm of scientific hypotheses. "All evolutionary hypotheses lead to a flagellum" is easily falsified. "Evolutionary hypothesis X leads to a flagellum" is also falsifiable.

The repeated failure to falsify evolutionary hypotheses A-M does not mean that hypothesis N will also fail, and for every hypothesis N, there is some hypothesis N+1. That's the definition of infinity. One cannot categorically rule out every evolutionary hypothesis.

Dembski claims that we can do just that, but he's wrong.

He claims that, if you compute the probability of a thing happening and it's less than the universal probability bound, you've got "design."

He gets that number by arguing that that is the maximum number of physical events that have happened since the big bang. If indeed that's true, it's easy to show that nothing could possibly have a lower probability. Any particle in the universe is, by definition, within this number of moves of any other particle. If a computation puts the probability of getting a certain result at less than the bound, that probability calculation must be wrong, by the definition of the probability bound. Most likely, one made inappropriate uniformity assumptions, assuming that each event was equally likely, when physical constraints made some outcomes more likely. That's exactly what happens in natural selection. Certain combinations are impossible, or of extremely low probability, while others are more probable.

It's worth noting that, to my knowledge, Dembski never actually attempts to compute a probability and compare it to this bound.

In fact, to run Dembski's test, you'd have to choose some small probability larger than the bound and show that some event is less probable than expected. Then we're dealing with a normal statistical test, one with two kinds of error.

Imagine a test to see if you are sick. There are four possibilities:
  1. You are sick and the test identifies you as sick.
  2. You are not sick, but the test identifies you as sick (Type I error).
  3. You are sick, but the test identifies you as not sick (Type II error).
  4. You are not sick, and the test identifies you as not sick.

Each is possible. You set the sensitivity of the test to minimize both types of error, though you may leave it too sensitive if you want to make sure you get every sick case, even if it means telling some healthy people they're well. But, short of telling everyone they're sick, you still have some non-zero probability of telling sick people they're healthy (Type II error). You also can't eliminate Type I error without telling all your subjects that they're healthy.

Similarly, Dembski cannot set a probability so low that only designed objects could possibly have that probability. No matter how low he gets (provided it's above 0) there will be a chance that an object which would be difficult to assemble by chance will, nonetheless, assemble by chance. The universe is very old and there are a lot of particles. Things happen.

IDC makes no falsifiable predictions about what they think will happen. It's vacuous and meaningless. If people didn't try so hard to take it seriously, I'd think it was a straw man.
But there four leading IDolators are, stating plainly that an unfalsifiable hypothesis is, in fact, falsifiable. That a hypothesis which fails to address design at all is in fact a design hypothesis. In short, that letter, sent to Science Magazine as what we have to assume wasn't a joke, is all one needs to invalidate IDC as a science.