On Design

Well, since Michael Egnor has sort of answered my questions, it's time for me to try to answer his. I'll try to be less evasive than he was.

One thing I'd like to point out is that Egnor seems to be under the misapprehension that the information theory that mathematicians and computer scientists actually study has something to do with inferring design.* This is simply not the case. Open up, for example, the book on Kolmogorov complexity by my colleague Ming Li, and you won't find a word about inferring design. (It's ID advocate Bill Dembski, of course, who is largely responsible for this confusion.) So, contrary to what Egnor thinks, as a mathematician and computer scientist I have no particular expertise on the general topic of "inferring design". It's just not something we do; maybe he should ask a SETI researcher, or a forensic investigator. But then again, Egnor has no particular expertise on the topic, either.

First, some general remarks about "design". I'll start by saying that I don't know exactly what he means by "designed". One of the favorite games of ID advocates is equivocation, so it's important to pin them down on a precise meaning. ID advocates rarely say plainly what they mean by "design". Do they mean simply that something has a pattern to it (as in "the design of a snowflake"), or do they mean something that has a "function", or must there necessarily be some teleology involved? I think it's incumbent on ID advocates to make clear what they mean. But I'll look at all three possibilities.

If ID advocates intend the first meaning ("pattern"), then the answer is clear: design need not imply a designer. The world is full of patterns that arise from the constrained nature of the physical universe. We can explain the overall pattern in a snowflake, for example, by referring to the symmetry in a water molecule in ice combined with the homogeneity of conditions as the snowflake forms. We don't normally bring in supernatural beings to explain snowflakes.

ID advocates, such as Bill Dembski, have nevertheless tried to rescue this hopeless case by bringing in probability. Dembski has argued that if an observed event fits a pattern, and the probability of this fit is extremely small, then the event must be due to design by an intelligent being. But as Wesley Elsberry and I have shown, his claim is based on incorrect mathematics combined with specious arguments. Dembski likes to say something on the order of 'witnessing specified events of low probability implicates design'. But the correct claim is merely 'specified events of low probability are never witnessed at all; if they are, that is prima facie evidence that your probability estimates are wildly off'.

Here's an example: there's a fellow with a web page who claims to have witnessed at least three independent meteorite falls. (The pictures he shows aren't meteorites, but let's ignore that for a moment.)

Now this page suggests that the probability of witnessing a meteorite land near you during your lifetime is about 1 in 2.5 million. So if the events are independent, we conclude that the probability of a particular person witnessing 3 meteorite falls in their lifetime is about 1 in 16 million million million. Even taking into account the total population of the world, this fellow's claim seems extremely unlikely. Could it have happened? Yes, but if so, we would have to consider some other possibilities: maybe the distribution of meteorite falls is extremely uneven, so that many more meteorites fall at this fellow's location than others. Maybe somebody's having him on, shooting meteorites out of a cannon towards his house. Maybe there's some other explanation entirely. In either case it's not that a specified event of low probability was witnessed; it was that our probability estimates were wrong. Dembski says design must be inferred when all other explanations are ruled out, but if my analysis is followed, inference to design no longer has the privileged place that Dembski accords it.

The second possible meaning of "design" corresponds to the inferred function of some object; this is basically the old argument of William Paley involving the watch found on the heath. While this argument may have been convincing two hundred years ago, it's convincing no longer, for the obvious reason that we know that evolutionary processes can produce function. We have good experimental evidence of this from novel mutations that, for example, allow some bacteria to digest nylon byproducts. As a computer scientist, I must also cite the artificial life experiments of Karl Sims, who showed how nontrivial and novel behaviors could evolve through mutation and selection -- something Dembski claims is impossible. Dembski has never addressed Sims' work.

The third possible meaning of "design" involves teleology; we infer a designer when we see something designed for a purpose. But either this begs the question, or it reduces to the previous paragraph about "function". So in all three cases, I don't think that seeing "design" implies a "designer".

ID advocates are always accusing others of 'wanting to eliminate the design inference from science'. Of course, this is pure nonsense. Archaeologists, for example, routinely attempt to deduce the roles that various objects played in the lives of the cultures they study. But, as Elsberry and Wilkins point out in their article from Biology and Philosophy, there is a huge difference between inferring design based on artifacts for which we have a causal story like human construction, and inferring design based on some causal story lacking any details whatsoever. They refer to this latter attempt, commonly used by ID advocates, as "rarefied design", and characterize it as "based on an inference from ignorance, both of the possible causes of regularities [that might explain the event] and of the nature of the designer."

One more point: I don't think that the question "is it designed?", in the absence of any candidate for a designer, is particularly interesting. That is, in the absence of motive, I don't think that knowing that something is designed tells you anything at all. I can do no better than to quote from Elsberry and Wilkins, who say

The problem with a simple conclusion that something is designed, is its lack of informativeness. If you tell me that skirnobs are designed but nothing else about them, then how much do I actually know about skirnobs? Of a single skirnob, what can I say? Unless I already know a fair bit about the aims and intentions of skirnob designers, nothing is added to my knowledge of skirnobs by saying that it is designed. I do not know if a skirnob is a good skirnob, fulfilling the design criteria for skirnobs, or not. I do not know how typical that skirnob is of skirnobs in general, or what any of the properties of skirnobs are. I may as well say that skirnobs are "gzorply muffnordled", for all it tells me. But if I know the nature of the designer, or of the class of things the designer is a member of, then I know something about skirnobs, and I can make some inductive generalizations to the properties of other skirnobs.

Now to Egnor's question. He wants to know why SETI is different from deducing design in biological systems. Referring to the fictional movie, Contact, where scientists received a blueprint for constructing a mechanical device, he asks, "If the scientific discovery of a ‘blueprint’ would justify the design inference, then why is it unreasonable to infer that the genetic code was designed?"

The answer is that I don't think that these situations are at all comparable. In the case of SETI, the fact that we are receiving a narrow-band signal is already suggestive, since we don't currently know any simple physical process that could produce these signals. This isn't a definite conclusion, though, because we have no idea what the probability of intelligent beings is, and we can't rule out narrow-band signals arising from some other physical process we simply don't know about.

In SETI, we are specifically looking for intelligent beings. These beings presumably live in another part of our universe, and presumably they evolved through natural processes, in much the way we did. This being the case, we hypothesize that, like us, these beings are interested in contacting other intelligent life, and would do so through radio waves. All these are assumptions based on our characterization of the "personality", if you will, of the originators of the signals. If any of these guesses are wrong, or if we are alone, we won't succeed. Our argument is based on analogy with our own thought processes, not "specified complexity" or other ID nonsense.

ID advocates, however, rule out any deductions based on the identity of the Designer. Yet, in real science, questions about intention and identity arise all the time in archaeology. To give just one example: in the 1890's, historian Arthur Evans heard of mysterious seal-stones from Crete. The identity of their creators, as well as the script used, was then unknown. Evans went on to identify the stones as the product of a civilization now called Minoan, and eventually one of the scripts, Linear B, was deciphered. The fact that ID advocates refuse to consider the really interesting scientific questions "who designed it?" and "why did they design it?" shows that they're not doing science.

To say that SETI is like the genetic code means that we have to hypothesize some designer who designed something for some reason. But where's the designer? In SETI, we can pinpoint a place in the universe where the signals are originating from. If the signals encode a machine, we can reasonably deduce that the intention is that we are to build it. But in the genetic code, who is the hypothesized designer? Where did they originate? When did they carry out their design? What is the intention of the design? All the really interesting questions are ruled as 'out of bounds' by ID advocates. Until they really come to grips with these questions, they're doing religion, not science.

As an example of something I'd find convincing, if we were to find a crashed spaceship with plans showing how to build a bacterium, and scientists carried out these plans and found that they really did construct life, then I'd find this very strong evidence that life on earth was designed.

Another point of disanalogy is that we know that DNA changes and evolves through time by processes such as mutation, gene duplication, and selection. There are even some very tentative answers to how the genetic code evolved. So the alternative hawked by ID advocates, that 'somebody made this sometime for some purpose, but we don't know where, when, or how', is not very impressive. In exactly the same way, a theory that lightning is 'caused by some intelligent being, but we don't know exactly how or why' is not very impressive, either.

Finally, DNA doesn't carry any of the hallmarks of human design, the kind of design we are most familiar with. Genes, for example, are often pleiotropic; they have multiple interacting effects. Human design, on the other hand tends to separate systems so they don't interact. Human activities tend to produce texts that are quite compressible; but a typical genome is hardly compressible at all. Biological entities reproduce themselves, but few, if any, human designs have this property. When we consider an analogy, like the one Egnor proposes, to be fair, we have to consider points of disanalogy, too. (For more examples of disanalogy, see Mark Isaak's article in Reports of the NCSE, Volume 23, no. 5-6.)

Let's alter the Contact story. Suppose the signal didn't encode a machine, but rather a sequence of DNA bases S. When we create DNA corresponding to this sequence, and stick it in a cell, we get an organism that tells us all about life on some other planet. Now the analogy is even closer than before; yet I think it is clear that our inference about the origin of S is still different from any inference about our own DNA. Indeed, it is entirely reasonable and scientific to infer that S is designed by intelligent beings on another planet, but our own DNA evolved.

Finally, on an unrelated note, Egnor takes me to task for my lack of civility. This is pretty rich, considering that Egnor's buddies at the Discovery Institute routinely insult the appearance of scientists, call them dishonest, and play games such as adding fart noises to Judge Jones' Dover decision. I say, clean your own stable first. Scientists are angry at the constant misrepresentation and juvenile antics of ID advocates, and we're not going to take it any more.

* Addendum: Because ID advocates have a track record of misunderstanding even the most trivial point, I should add that of course information theory could be used to separate, for example, human-generated text from noise. As I already mentioned, natural language text is quite compressible, while noise would probably not be compressible. But here we are not detecting design per se as some abstract category; rather we are using an empirical distinction between two things we have observed: natural language text, and noise.