No Free Lunch: Why Specified Complexity Cannot Be Purchased without Intelligence 
William A. Dembski 
Rowman & Littlefield, $35 (cloth)

"All science, even the divine science, is a sublime detective story. Only it is not set to detect why a man is dead; but the darker secret of why he is alive." 
                                                                                                                                                                                                                                           —G. K. Chesterton 1

Nothing evolves as surely as anti-evolutionism. The anti-Darwin movement, at least in its popular form, began in the primitive whoops and hollers of young-earthers and seven-day literalists. Their claims, as you might guess, were short on science and long on Genesis. But somewhat higher in the strata we find a thoroughly transformed, though recognizably related, beast: the scientific creationist. While still relying on some scriptural sources (many believed the fossil record reflected a certain forty-day deluge), these creatures did talk science, disputing radioactive dating and making lots of interesting claims about hydrology, pH, and sedimentation. Following their extinction, the strata reveal yet another and far more advanced form, the Intelligent Design champion. Compared to this modern species, its predecessors look downright primordial. Indeed the Intelligent Design advocate is characterized by at least three novel traits: i) advanced academic degrees; ii) sophisticated arguments accompanied by expert knowledge; and iii) strict avoidance of religious language, including any speculation on just who the designer might be.

Nothing evolves as surely as anti-evolutionism.

While usually admitting that life on earth is billions of years old and that people, pigs, and petunias are related by common descent, the Intelligent Design (ID) movement maintains that some bits of biology show the unmistakable handiwork of an intelligent agent. And while this agent may not wholly displace Darwin, the two at least stand shoulder to shoulder. The ID movement further maintains that intelligent design, as a legitimate scientific hypothesis, deserves a place alongside blind evolution in public schools and that students should, at the least, be exposed to both sides of the debate. Indeed Ohio, which is revising its curricular standards, is currently embroiled in a dispute over the possible introduction of intelligent design into its biology classes. (Texas, which dominates the U.S. textbook market, is gearing up for a similar dispute next year.)2 The ID movement is led by four tireless academics or ex-academics: Michael Behe (professor of biochemistry at Lehigh University), Jonathan Wells (biologist and senior fellow at the Discovery Institute, a Seattle think tank concerned with the "renewal of science and culture"), Phillip Johnson (professor emeritus of law at Berkeley), and William Dembski (associate research professor in the conceptual foundations of science at Baylor University and senior fellow at the Discovery Institute).3

Dembski—whose new book, No Free Lunch, is sure to ignite new firestorms over design vs. Darwin—is perhaps the most impressively credentialed of the lot. He wields a Ph.D. in mathematics from the University of Chicago, another in philosophy from the University of Illinois at Chicago, and a Master of Divinity degree from Princeton Theological Seminary. He is also author of seven books, including The Design Inference, a fairly technical work that laid out a statistical method allegedly allowing reliable detection of design.4 He is also an able writer, a skilled polemicist, and an indisputably bold thinker. And, yes, he believes—contrary to everything biologists told us for the last 150 years—that an intelligent agent helped shaped you and me.

To appreciate the magnitude of Dembski's claims in No Free Lunch you need to appreciate the relative modesty of Darwin's claims in the Origin of Species. Darwin did not rule out the formal possibility of a designer. Instead, he showed that the (apparent) design residing in organisms could be explained naturally, without recourse to a designer. And while he marshaled great masses of evidence for the role of his natural mechanism (natural selection) and against the role of a designer, Darwin made no claims about the impossibility of the latter hypothesis. Dembski's claims are more ambitious. Darwinism, he says, is formally incapable of explaining certain features of organisms. This is not to say that Darwinian mechanisms might not act now and then—Dembski agrees they might—but it is to say that Darwinism is mathematically barred from explaining certain things we always thought it could explain. And unfortunately for evolutionary biology, these things are not trivial arcana but the characteristic features of organisms: their staggeringly complex designs. (We'll sharpen the sense of "complex" below.) Dembski does not mince words: "[I]ntelligent design utterly rejects natural selection as a creative force capable of bringing about the specified complexity we see in organisms."

This is a big claim. And it explains why Dembski gets so much attention. You might whip up a bit of applause if you say that a designer can explain biology. But you'll bring down the house if you say that Darwinism can't and only a designer can. Especially if this claim gets dressed up in fancy mathematics of the sort that presumably intimidates biologists but snows the general reader. And this is precisely how Dembski dresses his claims. Borrowing results from computing theory—the so-called No Free Lunch theorems—Dembski claims to prove that Darwinism is utterly impotent before organismic complexity. Hence a designer. Unfortunately, Dembski's proof has nothing whatsoever to do with Darwinism and his claim to the contrary is hopelessly silly.

To show this, I need to back up and do two things. First, explain what kind of biological complexity Dembski is so worked up about and, second, explain why he thinks the No Free Lunch theorems stand in the way of Darwinism accounting for it. Doing this will require getting slightly technical for a moment. But don't worry—things will get simple again quick.

• • •

No free lunch

Not all complexity is a thumb in the eye of Darwinism. The problem, Dembski tells us, comes from a particular variety he calls "specified complexity":

An object, event, or structure exhibits specified complexity if it is both complex (i.e., one of many live possibilities) and specified (i.e., displays an independently given pattern). A long sequence of randomly strewn Scrabble pieces is complex without being specified. A short sequence spelling the word "the" is specified without being complex. A sequence corresponding to a Shakespearean sonnet is both complex and specified.

Dembski argues that biology is replete with specified complexity. It is certainly true that organisms are fantastically complex. It is also true that in some ways (but not others—this will become an issue) they are specified. It is clear for instance that the various parts of an organism are fitted to each other: the curvature of the lens is fitted to the distance to the retina so as to produce a sharp image. Dembski spends a great deal of time formalizing specified complexity in the language of information theory. Roughly speaking, we know we have a case of complex specified information if out of all possible ways of putting together a set of elements—say, all possible sequences of a set of letters and blank spaces—only a small subset represents a prespecified target and the actual outcome belongs to this target. Meaningful English phrases, for instance, represent a small target: the overwhelming majority of random combinations of English letters and blank spaces yield gibberish. So if you see a meaningful phrase (as you hopefully are now), you're seeing complex specified information.5

You might whip up a bit of applause if you say that a designer can explain biology. But you'll bring down the house if you say that Darwinism can't and only a designer can.

Now it's obvious how we go about making meaningful phrases: we use intelligence and crank them out at will. But how do biologists explain the complexity that resides in organisms? By Darwinism. To get a feel for what this means, consider the following caricature of Darwinism offered by Richard Dawkins and discussed at length by Dembski. Our target will be Hamlet's line, METHINKS IT IS LIKE A WEASEL. (Real evolution occurs in a sequence space that uses the four DNA "letters" A, G, C, and T but this is a distinction that doesn't make a difference.) First consider the odds of forming this target sequence by blind chance, i.e., with monkeys at word-processors. Draw a random letter from the alphabet for the first position in the phrase; now another for the second position, and so on. The odds that you've spelled out the phrase METHINKS… are essentially nil: in fact, with twenty-six letters plus a blank space, the odds of getting the word METHINKS alone are already less than one in 280 billion. But now consider the following "evolutionary algorithm." Start with a random sequence as before but i) randomly change each character that doesn't match the target sequence; ii) if a resulting character matches the target keep it and in the next round change only those characters that don't match. So, if we start with SATHINKS, at the next step we'll randomly change only the first two letters; and if those changes yield MQTHINKS, then at the next step we'll randomly change only the second letter. This two-step evolutionary algorithm of mutation plus selection arrives at the phrase METHINKS… with surprising speed.

This example also illustrates the idea of a fitness function. Fitness is a measure of quality; high fitness is good and low is bad. (In biology the only kind of quality that matters is how good you are at having kids. High fitness means you have a lot of kids and low means you have few.) A fitness function is just a mathematical function that assigns a fitness value to each possible sequence. In our Hamlet example, the best sequence is the phrase METHINKS…, so the fitness function assigns it the highest value. A sequence that matches METHINKS… at every position but one gets a slightly lower fitness, and one that matches METHINKS… at every position but two gets a yet lower fitness, and so on. A random sequence typically suffers a quite low fitness. If we now pretend that all possible sequences sit in a plane, we could plot their corresponding fitness values above this plane, forming a 3-D plot.6Evolutionists thus sometimes speak of fitness "surfaces" or "landscapes." Because evolution always moves from a sequence to another having higher fitness, natural selection can be thought of as moving populations uphill on fitness surfaces. In Dawkins's example this process ultimately arrives at the sequence METHINKS…, which sits atop a fitness peak.

Dembski's chief argument is that Dawkins's algorithm—and Darwinism generally—does not do what it seems. Indeed despite our unerring arrival at METHINKS…, the "Darwinian mechanism does not generate actual specified complexity but only its appearance." How can Dembski possibly claim such a thing? Enter the No Free Lunch theorems.

The NFL theorems compare the efficiency of evolutionary algorithms; roughly speaking, they ask how often different search algorithms reach a target within some number of steps.7Because the NFL theorems are deeply counterintuitive, it'll help to start with an informal rendition. It runs like this: If algorithm A beats algorithm B at some class of problems there will always be another class of problems at which B beats A. Further, one can show that A and B are equally efficient when averaging over all possible problems. The NFL theorems thus show that there's no such thing as a universally efficient algorithm: when faced with all problems, any algorithm is as good as any other. To appreciate Dembski's "generic" form of the NFL theorems, you need to appreciate that reaching a prespecified target with a particular fitness function is an example of a problem. Reaching the target with a different fitness function is a different problem. The NFL theorems thus say that if we average over all possible fitness functions—where some lead directly uphill to the target and others don't, and some are smooth and others rugged—no evolutionary algorithm outperforms any other. But one allowable algorithm is blind search, where we randomly move to a neighboring sequence regardless of its fitness (remember our monkey with a word-processor). The NFL theorems thus prove that no evolutionary algorithm beats blind search when averaging over all fitness functions. A surprising result.

The apparent success of Dawkins's algorithm at getting to METHINKS… must therefore be just that, an appearance. If Dawkins tried reaching his target when averaging over all fitness functions, he'd find he does no better than blind search. So why does Dawkins's algorithm seem to work? The answer is that it subtly cheats: it starts not only with a target but also with a fitness function that leads straight to it. Everything's been cooked into the fitness function. Algorithms like Dawkins's thus "fail to generate specified complexity because they smuggle it in during construction of the fitness function."8

Hence Dembski's big claim: "Darwinian mechanisms of any kind, whether in nature or in silico, are in principle incapable of generating specified complexity." At best, Darwinism just shuffles around preexisting specified complexity, using up that available in the fitness function to give the appearance of producing it de novo.

We can now complete the Dembskian Syllogism: Organisms show specified complexity; Darwinism can't make it; therefore, something else does. You won't be surprised to learn that that something else is intelligence. Indeed the "great myth of contemporary evolutionary biology is that the information needed to explain complex biological structures can be purchased without intelligence."

• • •

Nice answer, wrong question

The problem with all this is so simple that I hate to bring it up. But here goes: Darwinism isn't trying to reach a prespecified target. Darwinism, I regret to report, is sheer cold demographics. Darwinism says that my sequence has more kids than your sequence and so my sequence gets common and yours gets rare. If there's another sequence out there that has more kids than mine, it'll displace me. But there's no pre-set target in this game. (Why would evolution care about a pre-set place? Are we to believe that evolution is just inordinately fond of ATGGCAGGCAGT…?) Dembski can pick a prespecified target, average over all fitness functions, and show that no algorithm beats blind search until he's blue in the face. The calculation is irrelevant. Evolution isn't searching for anything and Darwinism is not therefore a search algorithm. The bottom line is not that the NFL theorems are wrong. They're not. The bottom line is that they ask the wrong question for what Dembski wants to do. More precisely, the proper conclusion isn't that the NFL theorems derail Darwinism. The proper conclusion is that evolutionary algorithms are flawed analogies for Darwinism.9

Dembski's chief argument is that Dawkins's algorithm—and Darwinism generally—does not do what it seems.

The astonishing thing is that Dembski knows all this. In a remarkable revelation—and one that follows two hundred pages of technical mumbo-jumbo—Dembski suddenly announces that Darwinists won't find his NFL objection terribly relevant. And why not? For the very reason I just gave. Dembski even quotes Richard Dawkins at length, who, it turns out, warned all along that his METHINKS… example is

…misleading in important ways. One of these is that, in each generation of selective "breeding," the mutant "progeny" phrases were judged according to the criterion of resemblance to a distant ideal target, the phrase METHINKS IT IS LIKE A WEASEL. Life isn't like that. Evolution has no long-term goal. There is no long-distance target, no final perfection to serve as a criterion for selection….In real life, the criterion for selection is always short-term, either simple survival or, more generally, reproductive success.10

At this point the reader of Dembski's book is a tad confused. Why, given the above revelation, is the book entitled No Free Lunch? Why is its dust jacket lined with blurbs from physicists attesting that Dembski has done something big? And, most important, why did I spend two nights reading about a theorem that reports an irrelevant result? The reader at this point has some right to know what Dembski's real problem with Darwinism is. And he comes through. After two hundred pages, Dembski finally unveils his Über-Objection: Darwinism does "not guarantee that anything interesting will happen." (I'm not making this up.) Darwinism, he admits, will work on a small scale—it will make bacteria resistant to antibiotics and insects resistant to insecticide—but it might not work on a big scale, yielding complex critters and the breathtaking biological diversity that envelops the earth. Dembski's problem isn't then with Darwinism per se. Like the scientific creationists before him, it's with Darwinism writ large. He's worried about the proper limits of extrapolation. And the non-extrapolationist evolution he ends up allowing—one that tinkers but doesn't innovate—is "certainly not a form of Darwinism that is worth spilling any ink over."

The Intelligent Design movement emerged out of a Judeo-Christian tradition that demands, or at least historically favors, an interventionist deity.

There are so many problems with this view that it's hard to know where to start. For one thing, it's wholly subjective. Though Dembski enjoys dressing up his claims in mathematical garb, his key objection to Darwinism ends up being a tad less rigorous than set theory: whether he finds the likely products of natural selection "interesting." For two of the 3.5 billion years of life, nothing fancier than bacteria lived on earth. Is this interesting? A virus might only have four genes. Is this interesting? Just where does one draw the line between beasts or changes that are sufficiently uninteresting that they can be subsumed under a Darwinian mechanism and those that are sufficiently interesting that they can't? Dembski's equations are silent here. For another thing, Dembski's anti-extrapolationist view leads him into some formal muddy waters. If, as he oddly continues to claim, the NFL theorems pose a problem for Darwinism, why don't they pose a problem for a little Darwinism? The NFL theorems don't say anything about scale. To say then, as Dembski does, that a little bit of Darwinism is okay (despite NFL) but a lot is bad (because of NFL) is to say something odd. Dembski comes precariously close here to saying that while there's no such thing as a free lunch, you can help yourself to brunch. Last, surely it's the refusal to extrapolate Darwinism from the small to the large scale that needs justifying. If Darwinism can explain small changes in organisms over the last fifty years (antibiotic resistance, say), surely it can explain progressively bigger changes over the last 500, 5000, or 50,000 years. The cumulative effects of mutation and selection aren't going to get smaller. Dembski's anti-extrapolationism seems a lot like saying that, while Kepler's laws might hold on any given day, they don't hold over whole years. Such a position is, I suppose, formally possible but it—and not extrapolation—requires special justification.

Alas, Dembski's attempts to explain why Darwinism won't extrapolate don't wash. He offers two reasons. The first is that things get simpler not fancier under Darwinism. "Simplicity by definition always entails a lower cost in raw materials…than increases in complexity, and so there is an inherent tendency in evolving systems for selection pressures to force such systems toward simplicity." Darwinism thus chokes when confronting a biological world that's so baroque. This is an ancient argument and the replies to it are equally old. Even if selection favors simplicity, note that the history of life must show a trend of increasing complexity. The reason is this history starts at zero complexity. On average it can only go up (where we cannot see the descendants of lineages that crashed and burned back into zero complexity). There are also good reasons for thinking that organisms get stuck at higher levels of complexity. John Maynard Smith and Eörs Szathmáry argue at book length that the formation of complex assemblies is often irreversible.11When free living mitochrondria and early cells came together, for instance, to make the first eukaryotic (true) cells, they swapped genes, so that mitochondrial proteins are now encoded by nuclear genes and vice-versa. At this point, things are essentially irreversible and the two partners can't go their separate, simpler ways. Dembski seems unaware of this well known point. Dembski's it-just-gets-simpler argument also relies on an erroneous assumption that natural selection cares primarily about the cost of raw materials. But selection cares only about how many kids you have. If I use more raw materials but have more kids than you, my type gets more common, period. Last, Dembski's argument is betrayed by his own examples of admitted Darwinism. When Salmonella evolved penicillin resistance and the mosquito Anopheles evolved DDT resistance just how did they get simpler? The answer is they didn't.12

Dembski's second anti-extrapolationist argument is that Darwinism could explain the fantastic range of biological diversity only if fitness functions are well-behaved. As he puts it, "the fitness function induced by differential survival and reproduction [may not be] sufficiently smooth for the Darwinian mechanism to drive large-scale biological evolution." If not, natural selection can't gradually ascend lofty fitness peaks and "there is no reason to think you will get anything interesting." Dembski tries here to reconnect his argument with the NFL world—you have to sneak in a fitness function that's just right. But the argument doesn't fly. To see this, consider fitness functions that are as unsmooth as you like, i.e., rugged ones, having lots of peaks and few long paths up high hills. (These are the best studied of all fitness landscapes.13) Now drop many geographically separate populations on these landscapes and let them evolve independently. Each will quickly get stuck atop a nearby peak. You might think then that Dembski's right; we don't get much that's interesting. But now change the environment. This shifts the landscape's topography: a sequence's fitness isn't cast in stone but depends on the environment it finds itself in. Each population may now find it's no longer at the best sequence and so can evolve somewhat even if the new landscape is still rugged. Different populations will go to different sequences as they live in different environments. Now repeat this for 3.5 billion years. Will this process yield interesting products? Will we get different looking beasts, living different kinds of lives? My guess is yes. Dembski's is no. And that is, I suppose, fine. He's entitled to his guess. But don't let him tell you that it follows ineluctably from some mathematical theorem because it doesn't. The troubling thing is that the above scenario isn't some contrived attempt to sidestep Dembski. It's the standard explanation of why organisms don't get permanently stuck on local peaks. For one brief moment Dembski seems to realize that changing environments might matter, pulling the rug out from under his it-won't-go-anywhere argument. But the worry is quickly dispatched with a footnote: "More precisely, f needs to be an evolving fitness function indexed by time. My argument, however, remains intact." Unfortunately it doesn't.

• • •

Irreducible complexity: once more with feeling

In the last half of his book, Dembski gets specific. He turns to an example of biological structures that is allegedly inaccessible to Darwinism. It would be more accurate to say he returns to the example as it's one that's been worked to death in ID circles. The idea comes from Michael Behe, the ID biochemist and author of Darwin's Black Box.14 Behe's argument was that some structures are "irreducibly complex": remove any part and the whole thing stops working. His favorite example was the mousetrap. Take away any part—spring or hammer, say—and function collapses. You won't catch mice. Behe claimed the biological cell is also loaded with irreducibly complex structures. His pet example, and one Dembski loves, was the bacterial flagellum, which sports a dizzying number of proteins that have to be arrayed in just the right way.

The importance of irreducibly complex structures is that they cannot, Behe assured us, be built by Darwinism. Darwinism demands that each step in the long walk to the present structure be functional. But that can't be: since all parts are required for function natural selection couldn't possibly have added them one at a time. Irreducible complexity is therefore a reliable marker of intelligent design. This argument sold a lot of books and got tremendous media airplay.

Unfortunately it was all wrong. Behe's claim was refuted—and in at least two ways. Both showed how irreducibly complex systems could be reached via gradual, Darwinian paths. Dembski calls the first path "scaffolding." At each step, a part gets added that improves a structure's function. At some point, however, a substructure might appear that no longer needs the remaining parts. These useless parts could then fall away. The key point is that the substructure we're left with might be irreducibly complex. Remove any part now and all hell breaks loose. The second path was one that I championed. Dembski calls it "incremental indispensability." Here's the argument:

An irreducibly complex system can be built gradually by adding parts that, while initially just advantageous, become—because of later changes—essential. The logic is very simple. Some part (A) initially does some job (and not very well, perhaps). Another part (B) later gets added because it helps A. This new part isn't essential, it merely improves things. But later on, A (or something else) may change in such a way that B now becomes indispensable. This process continues as further parts get folded into the system. And at the end of the day, many parts may all be required.15

Dembski more or less concedes that the above paths show that irreducibly complex machines can be built via Darwinism.16 Despite this, however, he bizarrely concludes that "[t]he challenge of irreducible complexity to Darwinian evolution is real, and to claim that Behe's ideas have been refuted is false." I must admit that I re-read this sentence four or five times, searching for signs it reflected multiple typos. But concluding that Dembski meant what he said, I tried to piece together why he still thinks irreducible complexity is a bone in the throat of Darwinism.

The answer is "causal specificity." The scaffolding and incremental indispensability arguments are not, Dembski says, causally specific. This means they have not, in any particular biological example, been fleshed out in sufficiently gory detail that Dembski can judge their validity. You might think scaffolding, say, can account for the bacterial flagellum but no one has told Dembski just which protein came first and which second:

Indeed, there is no way to argue against a putative transmutation that seems plausible enough to our imaginations but has yet to be concretely specified….This is of course another way of saying that the scaffolding objection has yet to demonstrate causal specificity when applied to actual irreducibly complex biochemical systems. The absence of detailed models in the biological literature that employ scaffoldings to generate irreducibly complex biochemical systems is therefore reason to be skeptical of such models.

This argument is more than a little annoying. Though Behe griped that evolutionists hadn't faced up to particular biochemical machines, his chief claim was that Darwinism just couldn't get here from there. He asked "What type of biological system could not be formed by 'numerous, successive, slight modifications'?" and answered "a system that is irreducibly complex." He announced that "[i]rreducibly complex systems are nasty roadblocks for Darwinian evolution" and spoke of "unbridgeable chasms." That's what all the hoopla was about, that's why Behe got in Newsweek, and that turned out to be dead wrong. So now the argument shifts. Now the problem is historical concreteness. But to leave readers with the vague impression that nothing's changed, Dembski brands his point "causal specificity." But this is a category mistake of the first magnitude. His point has nothing to do with causation. It's got to do with historical narrative. Which gene begat which protein in which order? Dembski's bait and switch here is transparent and puerile. If the ID community wishes to be taken seriously as honest intellectuals seeking truth (even if they're wrong; the two are not incompatible) they must plainly say: "Behe's chief claim was wrong. Irreducible complexity is accessible to Darwinism."

It is the height of hypocrisy for Dembski to complain that Darwinism lacks causal specificity when his own theory lacks any specificity, including one atom of historical concreteness.

The causal specificity argument is also an exercise in nerve. We are, recall, trying to choose between two theories. One says bacterial flagella were built by mutation and selection and the other says they were built by an intelligent designer. And Dembski concludes the first theory lacks historical concreteness? Darwinism suffers a shortage of specificity? When, after all, did Dembski's designer come up with plans for flagella? Just how did he reach out and shape that flagellum? Which protein did he move first or did he touch them all at once? It is the height of hypocrisy for Dembski to complain that Darwinism lacks causal specificity when his own theory lacks any specificity, including one atom of historical concreteness. Dembski may not have much of an argument, but you've got to admit he's got chutzpah.

Last, I can't help but wonder why Dembski's so worked up about irreducible complexity in the first place. Irreducibly complex systems do show specified complexity, but so do non-irreducibly complex ones. METHINKS IT IS LIKE A WEASEL is specifically complex (at least if it were longer) but it's not irreducibly so. So why the special treatment? Dembski seems to imply that irreducible complexity is special because it shows some structures can't be reached by smooth fitness functions. But this is refuted by scaffolding and incremental indispensability. The fact is that irreducible complexity plays no definable role in Dembski's view specifically and poses no challenge to Darwinism generally. The idea is dead and it's time the ID community gave it a proper burial.

• • •

IDing the designer

Dembski devotes some time at the close of his book to what ID as a practicing "science" might look like. This is one of the more interesting parts of the book. Dembski knows a fair amount about the history and philosophy of science and his observations here are on the whole worth hearing. It's also here that we learn Dembski's thoughts not on design, but the designer. Dembski considers two questions that reside in the No Man's Land between science and theology: Is the designer embodied or unembodied? And is design front-loaded in the universe (e.g., at the Big Bang and is now playing itself out) or periodically injected throughout cosmic history?17

Dembski's treatment of the second question is the more interesting as it leaves him in an especially awkward position. To be fair, Dembski admits that there are no grounds for excluding either front-loading or intervention. But it's clear where his heart lies. He seems less than crazy about the former idea and perceptibly leans to the latter. At the very least he defends intervention with gusto.18

What's odd about this is that Dembski goes out of his way here to make the slightest whiff of design maximally unpalatable to scientists. Plenty of scientists have, after all, been attracted to the notion that natural laws reflect (in some way that's necessarily poorly articulated) an intelligence or aesthetic sensibility. This is the religion of Einstein, who spoke of "the grandeur of reason incarnate in existence" and of the scientist's "religious feeling [that] takes the form of a rapturous amazement at the harmony of natural law." (This or something like it is also the religion of the young Chesterton with whom I began this essay.) This mild mysticism is fairly common among scientists, especially physicists and mathematicians. What's attractive about this view—which is of course thoroughly religious, not scientific—is that it at least requires no violation of methodological naturalism. The miraculous is not some alleged departure from natural law but the law itself.

Given that Dembski pays lip service to Duhem's claim that questions of coherence with existing theory invariably enter when choosing between views that explain the data equally well, you'd guess he'd rush to embrace Einsteinian front-loading. History shows it lives peaceably with science's remaining intellectual commitments. So why doesn't he? Why does Dembski work so hard to prop up interventionism?

I can only guess but the guess seems plain: Dembski's defense of interventionism reveals, I suspect, both the ID's movement's ideological roots and its political agenda. The movement emerged, after all, out of a Judeo-Christian tradition that demands, or at least historically favors, an interventionist deity. But more important, I suspect Dembski and much of the ID community are turned off by the fact that the Einsteinian view demands no change, much less revolution, in our practice of science. The Einsteinian view is insufficiently radical—too tame, too palatable, and too inconsequential for Dembski and his fellow travelers. It is one thing to stand in awe before the harmony of natural law. It is quite another to topple methodological naturalism, puncture materialism, and re-write the textbooks of Ohio and Texas. I can guess which Dembski prefers.



1 G. K. Chesterton, The Thing (New York: Dodd, Dead and Co., 1930), 72.

2 See Francis X. Clines, "Ohio Board Hears Debate on an Alternative to Darwinism," The New York Times, 12 March 2002. See also Trisha Gura, "Evolution critics seek role for unseen hand in education," Nature416 (2002): 250.

3 For links describing their publications, as well as those of other ID advocates, see the Discovery Institute on-line, For a critical analysis of the creationist/intelligent design movement, see Robert T. Pennock, Tower of Babel:The Evidence Against the New Creationism (Cambridge, Mass.: MIT Press, 1999). For a recent collection of papers defending and attacking intelligent design, see Pennock, Intelligent Design Creationism and Its Critics (Cambridge, Mass.: MIT Press, 2001).

4 The Design Inference: Eliminating Chance Through Small Probabilities (Cambridge: Cambridge University Press, 1998).

5 Strictly speaking, Dembski says we can infer complex specified information only if a phrase is long enough that the probability it would arise by chance falls below a "universal probability bound" of 10-150. So we'll assume throughout that target phrases are long.

6 Fitness landscapes are usually high dimensional, not three, but it's easiest, though not quite right, to imagine a 3-D landscape. Note also that the target evolution is shooting for needn't be a single sequence; it could include several. But, overall, the target is small.

7 These theorems were introduced by David H. Wolpert and William G. Macready, "No Free Lunch Theorems for Optimization," IEEE Transactions on Evolutionary Computation 1 (1997): 67–82. Dembski's "generic" form of the NFL theorem is loosely based on that of Joseph Culberson, "On the Futility of Blind Search: An Algorithmic View of 'No Free Lunch,'" Evolutionary Computation 6 (1998): 109–27.

8 To see that there's specified complexity in the fitness function, consider Dembski's further point: picking the right fitness function out of all those that are possible requires even more searching than picking the original target out of sequence space. So evolutionary algorithms just displace the task of finding a target back to the task of finding a desirable fitness function.

9 NFL theorems may well proscribe certain ways of talking about Darwinism (e.g., as a universally efficient optimizing algorithm) but that's a different matter. Dembski, incidentally, claims that "evolutionary algorithms…constitute the mathematical underpinnings of Darwinism" and that by "assimilating the Darwinian mechanism to evolutionary algorithms, [evolutionists] have invited a mathematical assessment of the power of the Darwinian mechanism to generate life's diversity." This is wrong. The mathematical underpinnings of Darwinism are population genetics, which does not consider pre-set targets and about which Dembski says nothing.

10 Richard Dawkins, The Blind Watchmaker (New York: W. W. Norton & Company, 1996), 50. Remarkably, Culberson—on whom Dembski leans for his interpretation of NFL—makes a similar point. Asking how biological evolution is possible given the NFL theorem, he speculates that perhaps "there is no global requirement on life other than it survive. Evolution was not necessarily looking for the human genome….We are not assuming the need for universal optimization, only very localized advantage." Culberson, "On the Futility of Blind Search," 123.

11 John Maynard Smith and Eörs Szathmáry, The Major Transitions in Evolution (Oxford: W. H. Freeman Spektrum, 1995).

12 In fact the evolution of antibiotic resistance often involves the gain of an extrachromosomal plasmid, i.e., an increase in the organism's total genome and, presumably, complexity.

13Stuart Kauffman and Simon Levin, "Towards a General Theory of Adaptive Walks on Rugged Landscapes," Journal of Theoretical Biology128 (1987): 11–45; John H. Gillespie, "Molecular Evolution Over the Mutational Landscape," Evolution 38 (1984): 1116–29; H. Allen Orr, "The Population Genetics of Adaptation: The Adaptation of DNA Sequences," Evolution (in press).

14 Michael J. Behe, Darwin's Black Box: The Biochemical Challenge to Evolution (New York: The Free Press, 1996).

15 H. Allen Orr, "Darwin v. Intelligent Design (Again)," Boston Review,December 1996/January 1997, 28–31. See also the exchange that followed in the February/March 1997 issue of Boston Review.

16 He says the "incremental indispensability objection is similar to the scaffolding and co-optation [which I skipped] objections in offering a narrative scheme for how an irreducibly complex system might conceivably have evolved by Darwinian means". And "[c]ertainly there is no logical impossibility that prevents such patchworks from forming irreducibly complex systems."

17 Dembski does not, though, consider another important question about the designer: What's gained by replacing a mysterious material order with an equally mysterious designer? This was one of Hume's objections to the argument from design. As Philo explains to Cleanthes in the Dialogues Concerning Natural Religion, "An ideal system, arranged of itself, without a precedent design, is not a whit more explicable than a material one which attains its order in a like manner; nor is there any more difficulty in the latter supposition than in the former." Cleanthes didn't have much of a response. It would have been interesting to hear Dembski's.

18 For a similar conclusion, see Robert T. Pennock, "The Wizards of ID,"Metaviews, 12 October 2000,