Dembski and Swamidass: What is the Status of the Explanatory Filter?

Dr. Dembski has responded to my recent interview: Axe and Swamidass: Should Christians Embrace Evolution? with an article at ENV:≠-repudiation

I stated in the interview:

Swamidass: “Dembski himself backed off from his book The Design Inference . He’s actually stated that he had it wrong in the explanatory filter. Are you aware of that?”

Axe: “He’s not backed off from the basic…”

Swamidass: “Yeah, I can show you the quotes later…. He’s even stated, I’ll give you the quote, that there was a gap in his argument. He doesn’t think the explanatory filter is the right way to make the ID case.”

Dembski responded, however, that I had him wrong.

In my book The Design Revolution , which appeared in 2004, six years after The Design Inference , I wrote:

Ultimately, what enables the filter to detect design is specified complexity. The Explanatory Filter provides a user-friendly way to establish specified complexity. For that reason, the only way to refute the Explanatory Filter is to show that specified complexity is an inadequate criterion for detecting design.

My position here hasn’t changed. I’ve beefed up specified complexity and developed it further over the years:


However, that is not the quote I was referencing. Nor did I claim that his retirement was evidence of his repudiation. In fact I agree, the issue is not his retirement nor whether or not he was repudiated. The issue instead is the status of the Explanatory Filter (EF). Here, I can’t reconcile Dembski’s article at ENV with this quote from 2008.

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." --Dembski, Vindication – The Austringer

In context, he is responding to criticism of the EF (Talk Reason: arguments against creationism, intelligent design, and religious apologetics). So that is important. He response to this legitimate critique was to back off from it, saying that he has dispensed with the EF any way. That’s just a fact.

It was good for him to do so, because EF has several logical flaws. See here:

The issue is that #1 and #2 are not an exhaustive list of all non-design causes. Until that gigantic loophole is closed, the EF is just not valid. This is important because (at least in principle) the EF can be invalid while the CSI is valid. By counter example, I’ve shown that there are large classes of entities that are not “designed” in this sense that come up positive. This logic seems to refute the EF in a straightforward way, and pointing to CSI is not a coherent defense for Dembski to offer.

Therefore, what Dembski says here is not accurate:

the only way to refute the Explanatory Filter is to show that specified complexity is an inadequate criterion for detecting design

In fact, on the surface, this claim is contradictory with his claim that he had “beefed up” the EF with his work in CSI. Why would CSI need effort to beef it up if the EF was adequate? In fact, it seems that any example produced by 4, 5, or 6 would be incorrectly called design when in fact they were not designed. THey would be “false positives” to the EF.

Now one objection often made is that all these false positives are in fact designed. After all, did not God create all things? And I agree. He did create all things. He designed Mount Everest after all! However, this objection is missing the sort of design that Dembski is after here in the EF. The structure of the filter is meant to detect design in a metaphysical context where some things are designed and others are not. If we rely on the doctrine that everything is designed…why exactly do we need to go through the song and dance of the EF any ways? Whatever the evidence, we can just say that God designed it.

(This is in fact is what Dembski is recognizing here: “[The EF] suggests that chance, necessity, and design are mutually exclusive. They are not.”)

So it does seem that Dembski did at one time back off the EF. It also seems that the EF has a massive logical loophole that has never been closed. It also seems that Dembski does not currently acknowledge any problems with the EF, but just redirects to CSI.

From here, it would be interesting to see how Dembski reconciles all this. It seems he had forgotten his quote where he backed off the EF. Can he contextualize it for us? Does he still think the EF is valid? If so, how does he deal with this logical error in the EF? Merely pointing to CSI isn’t enough, because the EF could be invalid even if CSI is valid. What exactly is the status of the EF in light of all this?


@Andrew_Loke I hope you’d be willing to offer your thoughts on this too. As a Christian philosopher, I know that this is an objection you’ve held too.

1 Like

Before design can be justifiably inferred, Josh argues, one must eliminate everything in (4), (5) and (6), below:

Thought experiment.

Question: what is the best explanation for the cause of this text? – i.e., the sentences you are reading right now.

Hypothesis: there exists an unknown natural cause, somewhere in (4), (5) and (6), other than the agent “Paul Nelson,” which produces English text about the epistemology of design inferences.

Dilemma: we cannot infer that the agent “Paul Nelson” wrote this text until we refute that unknown natural cause.

Therefore, the inference “Paul Nelson wrote this message” is invalid.

Hm. Something seems amiss.

Timely worry, Josh – I am writing a four-part series for ENV, “Trapped in the Naturalistic Parabola,” dealing with this very issue. First part should be up later today.

I’m sure we’ve talked about this before, but I think without satisfactory result. In what sense can it be said that Mount Everest (or perhaps Chomolungma?) was designed? Does the term “designed” become meaningless if we stretch it to include everything in the world?


This is why I think CSI is useless. You can describe a natural process that explains the CSI but then it’s proponent can just say that process was designed. They worked through that process. You can do that all the way back to the Big Bang.


That is not what I argued. I rather argued that the EF does not employ a valid inference.


In the context of this discussion, yes it does become meaningless. As I have stated elsewhere,

A Theological “Control” for Design

I am not an ID proponent, but I wonder if theology is the missing piece in the Intelligent Design Movement.

If we see evidence that evolution can do something surprising, we can just assert that this was due to God’s design. To study what biological systems can do on their own, we need to study biological systems operating on their own, without God’s design. If we mean God’s “design” to be intention and purposeful involvement towards a specific end, what are the biological changes He did not design?

It seems that this question is critical to answer, and it can only be answered by theology. Perhaps the evolution of viruses is one example of creation without design, and perhaps the evolution of cancer is another. Could these be “theological controls,” where biology can be studied without appealing to “design,” even if we believe God “created” all things? What other controls could we find?

I am not sure though how to settle these questions from science alone. It seems we need to engage deeply with theology.

This also why I think cancer is an important case study to test ID theories of innovation.

Of course, we know in fact that cancer can produce lots and lots of CSI…

1 Like


Because the first two nodes of the EF (natural law and chance) do not exhaust the universe of possible causes – your categories 4, 5, and 6.

So show me a case where design was legitimately inferred – say, plagiarism, archaeological discovery, Munchhausen’s by proxy – where the universe of 4, 5, and 6 was exhaustively searched.

I just did. I infer that Mt. Everest is designed, because God created all things.

Do you dispute my conclusion? Or do you dispute my logic? Or both? I hope not…


That’s because in your example we actually have the designer, Paul Nelson, and we know from experience that he’s generally in the business of writing english sentences.

A much better analogy would be when we have something we genuinely don’t know how came about, and we don’t have any currently known process (whether intelligent design, or unintelligent physical process) that could have plausibly given rise to it.

We just don’t have any good evidence of biochemists being around, say 2, or 3, or 4 billion years ago, willing polymers of amino acids and nucleotides into existence merely by thinking about it.

1 Like

Your understanding of “design,” as John Harshman points out, robs the notion of any empirical content. Everything is designed. And a skeptic would quietly add, to himself, “And therefore, nothing is, at least as far as science is concerned.”

You did not exhaustively search 4, 5, and 6. You declared, by fiat, that design is the case, come what may. Whatever that is, it’s not science, and Dembski and other ID people want to see how design can be rendered as a scientific hypothesis, with the explanatory power to discriminate events and causes.

I prefer to let my ENV series on the naturalistic parabola answer for me; I’ll post the link here when it’s ready. But let me end with a quote from one of Isaac Newton’s letters (to Burnet):

“Where natural causes are at hand, God uses them as instruments in his work, but I do not think them sufficient alone for the Creation.”

Why would Newton say that, if everything is designed?

I’m out until I have the new ENV link to post here.

That sounds like a problem. And if it’s a theological assertion that Everest was designed, theology is in trouble. If “designed” is meaningless, how can theology meaningfully use the term or assert its applicability? Surely even theology is bound by some sort of rules of sensible discourse.

Well, do we? Do we contend that God had the intention of producing the exact shape and composition of Mount Everest? How would he have been purposefully involved in producing the mountain? And is the same intention and purposeful involvement to be attributed in producing each particle of gravel or sand or clay? Certainly that would be within the capability of an omnipotent being (as would anything at all) but what purpose could possibly be served by such obsessive micromanagement?

To make matters worse, he also does not acknowledge any of the numerous problems with CSI that critics have pointed out over the years.


The focus here is on EF, not CSI. In principle, CSI could be valid, while EF is not. That’s important because appealing to CSI is not a defense of EF.

1 Like

But how does this apply to biology? People can go on and on about faces carved in mountains and words in a book, but at some point they have to get to biology.

After a very quick skim of “Algorithmic Specified Complexity”, the work on functional and mutual information done by Durston et al. seems to be tackling biology. This topic has been beaten to death in other threads, and I don’t think it would helpful to drag those dead horses out for a fresh beating at this very moment. However, this is probably the best entry point for a discussion, IMHO.

The other looming problem that others have alluded to is how we determine if a cause is natural, including those that happen right in front of us in the lab. ID is a big tent, and it includes concepts where a designer acts through natural causes. We would need a way to deal with these philosophical and metaphysical ideas before we could have a meaningful discussion.

The one glaring problem I see at the very root of this discussion is the Sharpshooter fallacy, and it is acknowledged in “Algorithmic Specified Complexity”

We would need to see how this could be avoided in the case of biology.


That isn’t what I did and not what I proposed @pnelson.

1 Like

Exactly this. It is why the ID project is doomed to failure, because all they could ever be doing is painting bullseyes around already existing gene-sequences we discover in living organisms.


Focus here is in EF, not CSI…

It would seem that the only way to answer that question would be a seance. Are you suggesting one?

1 Like

This is also a sign why ID has weird theological implications, in that some parts of nature are “less designed” than others. It just seems to fly in the face of more traditional conceptions of design and God’s role.