Dembski vs. Hume

William Dembski: ”The Design Revolution”. This is the latest book by William A. Dembski, author of “The Design Inference” and “No Free Lunch”.

The 32^nd chapter is available at Dembski's website.

The title of this chapter, “Hume, Reid, and Signs of Intelligence” indicates what it’s about: rebuttal of David Hume’s claim against design in nature in Dialogues Concerning Natural Religion, a book released 1779, three years after the death of Hume. Hume claimed that the appearance of design in nature was a weak argument from analogy, an inductive argument with no inferential chain to the supposed designer, wherefore, even if there was a designer, nature would be able to tell us nothing about that designer, not even if there was one or many.

The chapter can be divided into the following sections:

1. Defense of inference by analogy;
2. Hypothetical examples, where design should be inferred;
3. Thomas Reid’s argument against Hume; and
4. Specified complexity as the solution of the problem.

1. Defense of inference by analogy

Dembski admits that the inference of design in nature is based on analogy, but says, that where there is an analogy, there must also be a disanalogy, otherwise we would have an identity! So, how can we from knowing a watch to be designed conclude that an organism is designed? I will for this exposition rephrase Dembski’s argumentation. Let’s say we have two phenomena A and B. They share property p, that is p(A) and p(B). Additionally A has property q, that is q(A). Not knowing, if q(B), how can we then infer q(B)? Well, according to Dembski, we can make that inference, when p(X) => q(X) for all known X. This argumentation does not work, it clearly assumes, what we are supposed to prove, since B must be known. It’ll be clearer, if we look at Dembski’s example as he puts it in schematic form:

	P1	Watches are designed.
	P2	Watches and organisms exhibit functional interdependence of parts, adaptation of means to ends, etc.
	P3	There is no known instance where something exhibits functional interdependence of parts, adaptation of means to ends, etc. without being designed.
	C	Therefore, organisms are designed as well.

There is an obvious flaw in the argumentation here, it’s in the last premise (P3) and carries through to the conclusion (C).

The last premise should have read like this instead:

P3'

There is no known instance where something exhibits functional interdependence of parts, adaptation of means to ends, etc. without being designed or organic.

According to P2, organisms exhibit functional interdependence of parts, adaptation of means to ends, etc., and we do not have compelling evidence that organisms are designed.

It might be correct to claim that where we do have knowledge, namely in the case of watches and other clearly human-made products, we know that they are designed, but that does not allow us to extend the design to objects that are clearly very different. Otherwise we would allow for valid inferences by simply not investigating matters! We may tentatively suggest that organisms could be designed, but we cannot infer it, and we cannot reject attempts to explain organisms without assuming a designer. The mere fact that we have not yet fully understood abiogenesis, if it’s possible at all, does not allow us to refuse its possibility, certainly not as long as we cannot ourselves actually design organisms bottom up – and the day we can do that, if ever, that day we’ll have an explanation for abiogenesis anyway! In other words, the best indication that organisms are designed would depend upon our ability to design and manufacture them ourselves.

Let us replace “exhibit functional interdependence of parts, adaptation of means to ends, etc.” by complex. Then Dembski’s argumentation is:

	P1	(x) (watch(x) Þ designed(x))
	P2	(x) (watch(x) Þ complex(x)) and (x) (organism(x) Þ complex(x)).
	P3	Ø (Ex) (known(x) Ù complex(x) Ù Ødesigned(x)).
	C	organism(x) Þ designed(x).

It may be clearer here to see that the logic doesn’t work, because organisms are certainly known, although it is not known for sure, whether they are designed or not.

Dembski needs a stronger analogy than complexity, but he also claims to have one: specified complexity.

2. Hypothetical examples, where design should be inferred

But before coming to that, Dembski treats us to a few examples, where he assumes intelligence is applied, the first of which is based on the SETI (Search for Extra-Terrestrial Intelligence) project. This example is, however, actually based on the movie Contact from 1997 (which in turn was based on Carl Sagan's novel), where a long sequence of prime numbers was received, not on knowledge about the SETI project itself.

According to the SETI Institute (see SETI and Intelligent Design) the search is nothing like, what Dembski claims. Rather than searching for complex signals, the search is oriented towards simple signals (such as a single sinusoidal tone), but checks for Doppler shifts, which would indicate the source to be a planet spinning around itself or orbiting its local sun. The extra strange thing is that Dembski in his paper Is Intelligent Design Testable?, which is from 2001 (that is, long before The Design Revolution) himself mentions that

Although in the actual SETI program (Search for Extraterrestrial Intelligence) radio astronomers look not for something as flamboyant as prime numbers but something much more plebeian, namely, a narrow bandwidth of transmissions (as occur with human radio transmissions), the point nonetheless remains that SETI researchers would legitimately count a sequence of prime numbers (and less flamboyantly though just as assuredly a narrow bandwidth transmission) as positive evidence of extraterrestrial intelligence.

But if the SETI researchers are not looking for "something as flamboyant as prime numbers", how are they to know that there are any? The SETI researchers do not analyze radio signals for any kind of messages, so they won't register any prime numbers!

The other examples are no better and may leave you wondering, why Dembski doesn't at least try to come up with some examples that are somewhat realistic.

3. Thomas Reid’s argument against Hume

For Hume all knowledge about the external world is ultimately based on experience - there is no à priori knowledge. But, do we only detect design based on prior experience? How did we then detect design the first time around?

The Scottish philosopher Thomas Reid gave in 1780, four years after the death of Hume and one year after the release of “Dialogues Concerning Natural Religion”, (accidentally called “Dialogues Concerning Natural Selection” by Dembski, where he refers to Reid) a series of lectures on natural theology.

According to Dembski, Reid "demolished once and for all Humean induction as applied to design". In the lectures Reid remarked:

No man ever saw wisdom [read “design” or “intelligence”], and if he does not [infer wisdom] from the marks of it, he can form no conclusions respecting anything of his fellow creature. How should I know that any of this audience have understanding? It is only by the effects of it on their conduct and behavior, and this leads me to suppose that such behavior proceeds only from understanding. But says Hume, unless you know it by experience, you know nothing of it. If this is the case, I never could know it at all. Hence it appears that whoever maintains that there is no force in the argument from final causes [design], denies the existence of any intelligent being but himself. He has the same evidence for wisdom and intelligence in God as in a father or brother or a friend. He infers it in both from its effects and these effects he discovers in the one as well as the other.... From marks of wisdom and intelligence in effects, a wise and intelligent cause may be inferred. (Reprinted in Lectures on Natural Theology, University Press of America, 1981.)

Reid's claim here is that we do not infer wisdom (which Dembski translates into "design" and "intelligence") from seeing that wisdom directly, but by observing the marks of it. As William Paley would have put it, we infer from watch to watchmaker, not the other way around. And as Dembski might have put it, we infer from designed object to designer. We don't need to observe the designer nor enter the mind of the designer - we only need to observe the marks of design in the effects.

True, but how infallible is such a line of inference? Kepler claimed that the large craters on the moon were made by the inhabitants of the moon, because no natural processes could make such large round craters from a lot of small craters. However, what if the craters had been made by a single impact? The moon has no atmosphere for meteors to penetrate, so much larger meteor craters are possible on the moon than here on earth. Unfortunately, we don't have an ability, but rather a tendency to infer design.

4. Specified complexity as the solution of the problem

Dembski knows this quite well, so he comes up with specified complexity as the lithmus test for design.

As he puts it, a design inference is not an inference by analogy, but by isomorphy: specified complexity isn't only an attribute shared by all designed objects, it's an attribute shared only by designed objects.

But if that is so, how can design inferences then ever fail? Even a biologically founded tendency to pattern recognition may fail. That's how mimicry works in nature - by making pattern recognition fail.

An object exhibits specified complexity, if the object has a low probability to come around "by chance" (the "complexity" part), but conforms to a pattern that can be recognized independent of any knowledge of the causal history of the object (the "specified" part). But how is that different from jumping to conclusions? We don't know the causal history of some object, but it looks like something we know, so it must be something we know!

As for design and biology, in 1818 the 21 year old Mary Shelley wrote one of the classics of world literature, Frankenstein.