I’ve been saying that there were problems in William Dembski’s “explanatory filter” for a long, long time. I published a book review of The Design Inference back in 1999 that included the following:

According to Dembski, because humans identify human agency using the explanatory filter, the explanatory filter encapsulates our general method for detecting agency. Because TDI is equivalent to the explanatory filter, the conclusion of design in TDI is equivalent to concluding agency. Dembski specifies a triad of criteria — actualization-exclusion-specification — as sufficient for establishing that an intelligent agent has been at work, and finds that design as he uses it is congruent with these criteria.

However, Dembski’s triad of criteria for recognition of intelligent agents is also satisfied quite adequately by natural selection. “Actualization” occurs as heritable variation arises. “Exclusion” results as some heritable variations lead to differential reproductive success. “Specification” occurs as environmental conditions specify which variations are preferred. By my reading, biologists can embrace a conclusion of design for an event of biological origin and still attribute that event to the agency of natural selection.

John Wilkins and I took up criticism of Dembski’s “explanatory filter” in our 2001 peer-reviewed paper, The advantages of theft over toil: the design inference and arguing from ignorance, finding that Dembski’s supposedly fixed and mutually exclusive categories didn’t work so well when one took care in examining how he proposed to place instances in those categories.

Did Dr. Dembski thank me or us for getting that right? No, don’t be silly. But get it right we did, and there is an admission that the “explanatory filter” doesn’t work from William Dembski.

(1) I’ve pretty much dispensed with the EF. It suggests that chance, necessity, and design are mutually exclusive. They are not. Straight CSI is clearer as a criterion for design detection.

I don’t know that “straight CSI” offers any improvement; after all, CSI was what the “explanatory filter/design inference” was supposed to identify. But I guess when it comes to Dembski recognizing faults in his work, we will have to be satisfied with baby steps.

Wesley R. Elsberry

Falconer. Interdisciplinary researcher: biology and computer science. Data scientist in real estate and econometrics. Blogger. Speaker. Photographer. Husband. Christian. Activist.

18 thoughts on “Vindication

  • 2008/12/11 at 3:19 am

    Sorry to bust your bubble but Bill has reinstated his EF by popular demand.

    Seems the EF works after all.

    “It ain’t over till the fat lady………evolves”.

  • 2008/12/11 at 7:40 am

    First of all the explanatory filter is NOT mutually exclusive.

    Once one gets past a step the next strep includes all the processes that were before it plus the process at the present decision node.

    Anyone with any investigative experience would know that. However desk-jockies would not.

    And as a matter of fact the EF is used by people around the world.

    For example I was watching a show on finding bigfoot. It had scientists roaming the woods looking for signs. They came across something- felled trees set-up in such a way that one determined “design” almost instantly. When questioned by the other scientists the first said basically that necessity and chance cannot account for that formation.

    Next we have the “Ghost Hunters” on the SciFi channel. They use the EF- as in first they try to explain phenomenon X via all other non-agency explanations. Only after those attempts have failed, and a pattern is met, do they say the place is haunted.

    And the bottom-line is if you want to refute ID pure rhetoric and misplaced rhetoric at that, you have to provide some actual data that supports YOUR position.

    Whining at the other guys just doesn’t cut it. But I understand why you do that- because you can’t provide said data.

    As a matter of fact I doubt the “theory” of evolution canb even muster a hypothesis:

    For several years I have been asking for a testable hypothesis for evolution via an accumulation of random mutations- ie genetic accidents.

    The best I have seen so far only include hypotheses for descent with modification minus the mechanism.

    And in each case I was able to take the hypothesis and turn it so that it supported ID or convergent evolution. That is why a testable hypothesis for the stated mechanism is key.

    Any takers?

  • 2008/12/11 at 8:00 am

    Now I am curious- it not via the EF how have we (humans) been determining design for all these years?

    Archaeologists look for signs of work. Signs of work mean it is not explainable by necessity and/ or chance.

  • 2008/12/11 at 8:34 am

    It would seem that Dembski’s massive ego will not allow criticism:


    William Dembski


    12:59 am
    DaveScot: Right. I came up with the EF on observing example after example in which people were trying to sift among necessity, chance, and design to come up with the right explanation. The EF is what philosophers of science call a “rational reconstruction” — it takes pre-theoretic ordinary reasoning and attempts to give it logical precision. But what gets you to the design node in the EF is SC (specified complexity). So working with the EF or SC end up being interchangeable. In THE DESIGN OF LIFE (published 2007), I simply go with SC. In UNDERSTANDING INTELLIGENT DESIGN (published 2008), I go back to the EF. I was thinking of just sticking with SC in the future, but with critics crowing about the demise of the EF, I’ll make sure it stays in circulation.”

  • 2008/12/11 at 11:04 am


    Dembski said he would continue to promote the EF. Dembski did not address the problems inherent in it. There is a difference.


    You seem to be behind the times. I was pointing out problems in the EF in 1997. And Wilkins and I already addressed various of your EF concerns in our peer-reviewed paper. SC or CSI doesn’t help. Check out SAI in the appendix there.


    If there is something Dembski doesn’t tolerate well, it is criticism.

  • 2008/12/11 at 12:56 pm

    Sorry to bust your bubble but Bill has reinstated his EF by popular demand.

    Seems the EF works after all.

    Popular demand? Is that how you tell if something about nature is valid?

    A more ridiculous thing I’ve not read today.

    This is hilarious. So, if Dembski accepts the EF, it’s a powerful tool that works even when you don’t realize you’re using it – but if he rejects it, well of course we knew it never worked all along. The heads of the poor sycophants on UD must be spinning…please Bill, tell them what to believe and stick with it!!

    Dembski doesn’t have to say the EF is dead. The mere fact that it has never once been used for any real biological system, not even by Dembski, is plenty of evidence that it was never alive in the first place. The closest he got was the flagella, and even then he used a model divorced from reality and irrelevant.

  • 2008/12/11 at 2:11 pm


    How much investigative field experience do you have?

    And just because your article was peer-reviewed justmeans you found some people to agree with it.

    The EF works. The only “flaws” are 1) the data input and 2) the people/ person using it.

    And all anyone has to do to refute the design inference pertaining to biological organisms is to demonstrate that undirected processes can account for it.

    Can you even muster a hypothesis for the premise of undirected processes?

    Did you cover that in your paper?

  • 2008/12/11 at 2:18 pm

    One more question- How can you say out of one side of your mouth “We don’t know” (one of your outcomes) and out of the other say “But we know it wasn’t via intentional design”?

    And one more point- It is called a design INFERENCE for a reason- further research can either confirm or refute it.

    But the main point is we know it matters to an investigation whether or not that which is being investigated arose via agency involvement or nature, operating freely.

    And in the end nature operating freely cannot account for its own existence…

  • 2008/12/11 at 7:31 pm

    I’m always struck by the philosophical baggage which comes with the EF, particularly the assumption of dualism between mind and matter. I mean the EF takes a priori that the two are mutually exclusive “agency involvement OR nature”.

    If, in reality, intelligence is a subset, or product, of regularity (i.e. generated by a natural processes) the EF would collapse into a decision tree about either chance or types of regularity.

    Take a moment to drop the designer into the EF and the whole matter boils down to whether you prefer an infinite regression of increasingly complex designers, each one the progenitor of the last; or that, at some stage, significant complexity has originated from a form of ‘natural’ systems – contradicting a starting assumption of the EF.

  • 2008/12/11 at 11:38 pm

    Joe G.,

    You said,

    I wonder if these clowns think that all the success people have had using the EF or some reasonable fasimile thereof, is just an illusion?

    There are plenty of flaws to the EF, as my commentary starting from a few minutes after Dembski presented in 1997 through a paper currently in press point out.

    One big problem in discussion is that people often confuse, as you do, assertions with inference. The EF as given by Dembski in “The Design Inference” does not itself deliver a statement about agency. Dembski spends part of chapter 2 attempting to argue from his EF to claims about agency, but without result.

    Dembski’s claims for the EF, though, are not the modest sort of thing that is open to results of “further research”. Dembski claimed, and never retracted the claim, that once the “design” category of his EF was invoked, no further information could possibly overturn that result. It is a false claim on his part, as John Wilkins and I effectively pointed out.

    A rarefied design inference is not delivered by Dembski’s EF. Ordinary design inferences are something that John and I were clear can be made, just not via anything as error-ridden as Dembski’s EF. In the cases where Dembski’s urges people to make rarefied design inferences, we are pointing out that there is no justification or warrant for the claim. Where we know the properties, means, and motivations of designers, it is easy to justify an ordinary design inference based on that information, so I’d have to reject the misrepresentation that John and I have somehow rejected any form of detecting ordinary design.

    As for a proposal for making ordinary design inferences, Jeff Shallit and I took that up in our examination of CSI. Dembski mentions “rational reconstruction”, but his EF fails to be rational. The faults extend to Dembski’s CSI, the property that the EF was supposed to reliably identify. Jeff and I developed an application of the “universal distribution” that we called “specified anti-information” or SAI. This is an application of algorithmic information theory and identifies just those phenomena that are best explained as the result of a simple computational process. In the case of agents following effective methods, it does the useful work of a “rational reconstruction” that Dembski claimed, but did not deliver, for his EF. SAI is simple to apply to real world problems. Dembski’s EF has “resisted detailed applications to real-world problems”.

    As a matter of fact, nobody uses Dembski’s EF. Even Dembski has not done so for any non-trivial real world problem. Nobody goes to the trouble of following the full procedure given in “The Design Inference”. Nobody goes to the trouble of following the full procedure given in “No Free Lunch”. At best, you have people claiming that Dembski’s EF is in use when what they are actually talking about is what Jeff Shallit and I called “the sloppy chance elimination argument”. See also Gary Hurd’s excellent refutation of the canard that Dembski was describing how work gets done in archaeology, as a chapter in “How Intelligent Design Fails”.

    Something that seems a perennial feature of antievolution cheerleaders is this major dose of braggadocio that many bring to a discussion. Bob Schadewald referred to it as “the elevation of mediocrities”. There’s nothing to back it up other than a bully’s temperament, so once they get started, one can predictably watch as they deploy loads of ad hominems into the discussion while making an inflated pretense to expertise.

    Now, I should note that somebody coming into my living room and calling me a “clown” (or even stepping just outside and then shouting it) isn’t likely to be hanging out there long. So everybody wave good-bye to “Joe G.” Don’t bother coming back, Joe; the door is shut.

  • 2008/12/12 at 8:00 am

    Hey Joe…too bad you wewre invited out. Just when I was about to ask for some of those allegeded successes in using on EF on real biological systems. So far we have a few artificial contrived systems having nothing to do with biology and lot of handwaving. And those contrived examples are like the dowser who confidently smirks when his sticks bend over a glass of water. Sadly, he tends to do no better than chance when the water can’t be seen.

    And sorry Joe, but no amount of research can ever confirm or deny design as that term is used by Dembski. Design in his sense exists in the shadows of what we don’t know yet, and even if every single shadow is illuminated in one system, he can simply shift to another. And even if all the shadows are illuminated, there STILL exists the possibility that thats just the way the designer did it. You need the learn the difference between ‘doesn’t’ explain and ‘can’t’ explain. Creationists tend to equivocate between the two.

    Dembski design concoction is impotent as a scientific concept.

    Unless we consider natural selection an intelligent designer. Which if we use Dembski’s definition of ‘intelligence’ as chosing between options sans conciousness, we can.

  • 2008/12/12 at 10:24 am

    CSI is clearer? Wow he must be joking. He cannot give a definitive mathematical relation for what he means by CSI.

    Clearly, what he must mean, is the it is clear that he dodge objections using CSI better because it is so much less definitive.

  • 2008/12/12 at 10:26 am

    BTW – trying to apply the EF is hopeless. You cannot answer the first question right off the bat except when you actually do know of a natural law. So there is no way to get to the second test.

  • 2008/12/12 at 12:02 pm

    Yes, it’d be nice if specification was something more rigorous than ‘we know it when we see it’. That’s porn, not science.

  • 2008/12/14 at 2:50 pm

    I guess neither Joe G. nor Dembski read your 1999 book review.

    “Actualization” occurs as heritable variation arises. “Exclusion” results as some heritable variations lead to differential reproductive success. “Specification” occurs as environmental conditions specify which variations are preferred. By my reading, biologists can embrace a conclusion of design for an event of biological origin and still attribute that event to the agency of natural selection.

    I’ve never been able to understand how Dembski turns “what we found” into “Specified.” Who specified it?

  • 2009/06/04 at 11:38 am


    There is a brief passage in Dembski’s “No Free Lunch” that sheds some light on that. See my almost equally brief post at AE:

    William Dembski’s No Free Lunch contains the following passage:

    The presumption here is that if a subject S can figure out an independently given pattern to which E conforms (i.e., a detachable rejection region of probability less than alpha that includes E), then so could someone else. Indeed, the presumption is that someone else used that very same item of background knowledge — the one used by S to eliminate H — to bring about E in the first place.

    No Free Lunch, p. 75

    Because Dembski’s framework is based upon the elimination of alternative explanations, what we end up with here is the situation that Dembski is attributing the complement of the probability that can be assigned to chance hypotheses to an implicit design hypothesis, the one that underlies a particular “specification”. When the “saturated” probability of the alternative is less than 1/2, Dembski says that we should prefer “design” as our causal explanation, and because we have this relationship between the specification and the putative causal story, we thus are adopting that particular causal hypothesis.

  • Pingback: Beavers of the Gaps | The Questionable Authority

Comments are closed.