Another Round On Specified Complexity

There's a famous short story by Woody Allen called “The Gossage-Vardebedian Papers” that I like to reread from time to time. (It's very short, so follow the link if you've never read it before.) The story is told through the correspondence of Gossage and Vardebedian, as they argue about a game of postal chess in which they are engaged. There's one excerpt that keeps coming back to me, since it applies so perfectly in so many contexts:

Received your latest letter today, and while it was just shy of coherence, I think I can see where your bewilderment lies. From your enclosed diagram, it has become apparent to me that for the past six weeks we have been playing two completely different chess games—myself according to our correspondence, you more in keeping with the world as you would have it, rather than with any rational system of order.

I was reminded of that when reading the latest missive from Winston Ewert on the subject of specified complexity. It has become apparent to me that we have been engaged in two entirely different conversations--myself, according to what has actually been said, Ewert more in keeping with the world as he would have it.

For the background: Ewert's original essay is here. I then replied in two parts: here and here.

As an example of how things are going to go, here is Ewert's opening:

In a recent series of posts here at Evolution News, I answered objections to arguments for intelligent design based on specified complexity and conservation of information (see here, here, here, and here). The series, in turn, provoked responses that I would like to address now. Over at Panda's Thumb, University of Washington geneticist Joe Felsenstein declares that it is “Game over for antievolutionary No Free Lunch argument.”

But if you follow the link Ewert provides, you will find that it was Nick Matzke who declared “Game Over,” not Joe Felsenstein.

Moving on, we quickly come to this:

In both cases, the authors claim I've essentially admitted that the arguments behind specified complexity and conservation of information were incorrect. Both accordingly declare that the arguments we have made are now vanquished. But this is simply not true. What I've called false is the straw-man version of these arguments, which our critics chose to attack over the real ones. And I haven't “admitted” this so much as done my best, repeatedly, to clarify the difference between the straw man and genuine arguments for ID.

I'll let Joe speak for himself, but I certainly never claimed any such thing. I did not say Ewert essentially admitted that the arguments behind specified complexity were incorrect. What I actually said, after providing relevant quotes from Ewert, was this:

Ewert has simply conceded Felsenstein's point here. Assertions of specified complexity are entirely parasitic on prior arguments about irreducible complexity and the like. It is those prior arguments that are doing all the work, and not any subsequent claims about specified complexity. For example, if Behe's claims about irreducible complexity were correct, then that would all by itself be a very strong argument against evolution. No probability calculation would be needed to make it stronger. But since Behe's claims are not correct, no calculation based on the assumption that they are is going to be relevant.

That's the point. Assertions of specified complexity contribute nothing to the argument between evolution and ID. Ewert “effectively admitted” that by acknowledging that any relevant probability calculation must assume the correctness of prior ID arguments, such as Michael Behe's claims about irreducible complexity. It is those prior arguments that are doing all the work.

Ewert now continues:

Specified complexity has always required the calculation of probability. It has never been based on assuming that every outcome was equally probable.

After providing links to his previous writings he writes:

Yet Rosenhouse insists that specified complexity required calculation assuming every outcome is equally likely. For support, he points to a section of a critique of No Free Lunch by Richard Wein (“Not a Free Lunch but a Box of Chocolates”), describing this as an exhaustive documentation for his claim. But Wein himself describes the evidence as inconclusive. Wein points to examples and arguments that he thinks imply uniform probability, but he ignores the places where William Dembski explicitly states that the calculation has to be done according to the hypothesis under consideration.

First of all, I said Wein provided “extensive” documentation, not “exhaustive.”

More importantly, did I insist that specified complexity required calculation assuming every outcome is equally likely? I don't recall insisting on any such thing. The way I remember it, what I actually said was this:

[T]he problem here is that Dembski, in his various writings on this subject, is incredibly vague on how, precisely, one carries out the relevant probability calculations. Look hard enough and you can find him saying just about anything you want him to be saying. But as Richard Wein has documented extensively, in response to Dembski’s book No Free Lunch, Dembski routinely writes as though the probability calculations should always be carried out with respect to a uniform distribution, which is what Felsenstein had in mind in referring to chance alone.

My point was that Dembski is vague, and that you can find him saying just about anything you want him to be saying on the subject of probability calculations. I then pointed out that Dembski often writes as though a uniform distribution should be used. Indeed he does, as Richard Wein documented extensively. In some places Dembski seems to assume arbitrarily that a uniform probability distribution should be used, in other places he suggests some other method. Hence the frustration of his many critics in trying to pin down precisely what he is claiming.

As for specified complexity, my main point was this:

But in any case where we are genuinely uncertain as to whether the event or object is the result of design, we are also going to lack the information to carry out relevant probability calculations and to identify the design-suggesting patterns.

Why would I have said this, if I thought we were meant to assume a uniform distribution? In trying to calculate the probability of a biological structure, like a flagellum, the information we lack is precisely the correct distribution to apply to our space. If I thought we could just assume a uniform distribution, the calculation would be simple.

And that's what it all comes down to. There is no way to carry out a meaningful probability calculation in any biological case. That is hardly the only problem with the concept, but it is sufficient for dismissing it from the conversation.

Finally, we come to this:

In my most recent discussion of specified complexity, I quoted Rosenhouse, which prompted his responses. His primary objection is that I've quoted him out of context. I quoted his presentation of the argument that improbable events happen all the time, but did not reference his later discussion of specified complexity.

The fact is, Rosenhouse has misread the context in which I discussed him. I dealt there with a particular objection to specified complexity. Felsenstein had claimed the correct version of specified complexity was useless. It is not useless, because specified complexity justifies rejecting an explanation that assigns too low a probability to the outcome it purports to explain. If I win the lottery every day for a month, you are entitled to conclude that I was not playing fairly. The objection is raised that this is plainly obvious, and it is sometimes suggested that no reasonable person would ever raise this objection. But they have indeed done so.

In quoting Rosenhouse and the others, my purpose was not to make them look bad or to suggest they don't understand specified complexity. My point was to show that people do actually raise this objection. Critics point out that low probability events happen all the time, and that this undermines rejecting an explanation on improbability alone. This objection, a valuable one, was not a figment of Dembski's imagination. It's a real argument that needed to be addressed.

I misread the context? Actually, I did in one sense. Ewert doesn't say that I was the one disparaging specified complexity in that quote, so it was wrong of me, in my earlier post, to suggest that he had. Sorry about that!

But the rest of it is nonsense. First, Ewert introduced his quotation of me like this:

Some critics of intelligent design regard this as an obvious point. If complex life were prohibitively improbable under Darwinian evolution (an idea these critics certainly reject), Darwinian evolution would clearly be false. They find it difficult to believe that specified complexity was developed to defend such an obvious point. However, other critics insist that low probabilities of complex life would not provide evidence that we should reject Darwinian evolution.

I was one of three people charged with rejecting an obvious point. After presenting these quotations, Ewert writes:

All of these critics argue that we cannot draw conclusions about Darwinian evolution from small probabilities. To be fair, many critics would not agree with these simplistic criticisms of probability arguments. They would in fact argue that evolution makes complex biological systems highly probable. It is unfortunate that these writers feel the need to disparage specified complexity, which exists to defend against an argument they would not make.

My argument is described as simplistic, to the point where fairness to other ID critics requires acknowledging that many would reject what I am saying. So I think I can be forgiven for concluding that the point here was to make me look bad.

Substantively, Ewert is not making any sense here. He seems to think it is silly of me to argue that we cannot reject Darwinian evolution just from finding that one of its outcomes was highly improbable. Another person who makes this argument is William Dembski, as I documented in my previous post. And rightly so! Of course low probability by itself is never enough to reject an explanation.

This is clear even from Ewert's lottery example. Consider the last thirty winners of the lottery. The probability of precisely those thirty people winning those lotteries is exactly the same as the probability of one person winning the lottery thirty times (given some reasonable and obvious assumptions). But the first scenario is not suspicious while the second one is. Of course I would suspect foul play if one person wins the lottery thirty times in a row, but it would not be just because it is very improbable for one person to win the lottery thirty times.

Most bizarre, though, is his claim that specified complexity was meant to refute those benighted critics who think that low probability alone is insufficient for rejecting Darwinian evolution. The whole point of talking about “specified” complexity is that complexity (improbability) by itself is insufficient for rejecting an explanation other than design. If improbability were enough, you wouldn't need to talk about specification at all. The need for a notion like specification is the confirmation of what I, and Ewert's other examples of simple-minded critics, were saying.

Let's try to end on a substantive note. If Ewert believes that specified complexity has contributed something to the discussion between evolution and ID beyond the arguments made by other ID writers, then let him point to a specific example where that is the case. The entire body of ID literature records precisely one example where the machinery of specified complexity was applied to a specific case: William Dembski's flagellum calculation in No Free Lunch. That calculation was entirely parasitic on Michael Behe's arguments about irreducible complexity, and was ludicrous even granting, for the sake of argument, the correctness of those arguments. Since Behe's arguments were not correct (Dembski's attempted refinements of those arguments notwithstanding) it's unnecessary to pay any attention at all to the details of the calculation.

If Ewert has something better than that, I'm all ears. Since I suspect he does not, the rest of this is just posturing.

More like this

Over at the Discovery Institute's blog, Winston Ewert has a post up explaining, one more time, what specified complexity is. Since I am given a mention near the end, perhaps it's worth a look. For those not steeped in ID rhetoric, “specified complexity” is a term coined by William Dembski. It is…
Let us continue with our discussion of Winston Ewert's defense of the concept of “specified complexity.” In Part One we saw that Ewert's defense was actually rather tepid. He mostly gave away the game by writing: It is true that specified complexity does not in any way help establish that the…
I had not intended to do another post on this topic so soon after the last one. But I have just readan astonishingly bad post over at Uncommon Descent that discusses this issue, and I cannot resist responding. The post is called, “Where Do We Get the Probabilities?” It was written by Winston Ewert…
As it happens, I've been thinking about mathematical anti-evolutionism a lot lately. Sometime over the summer, though I can't find the exact post, I mentioned that I had been working on an article about mathematical arguments against evolution. I finished it in the fall, and it has recently been…

Ewert wrote:

But Wein himself describes the evidence as inconclusive. Wein points to examples and arguments that he thinks imply uniform probability, but he ignores the places where William Dembski explicitly states that the calculation has to be done according to the hypothesis under consideration.

As it happens, Wein is a regular reader of this blog, so I'm on hand to clarify just what I did and didn't say. Far from "ignor[ing] the places where William Dembski explicitly states that the calculation has to be done according to the hypothesis under consideration", I devoted a whole section of my article (section 3) and a lengthy appendix to Dembski's method of inferring design by calculating a probability under each "relevant" hypothesis. These contain many references to "P(E|H)" and "P(R|H)", i.e. the probability of an event given the hypothesis under consideration. The problem is that Dembski himself does not stick to this method. He frequently writes as if we need consider only one probability distribution, and in his examples this is almost invariably a uniform probability distribution, even when there are reasons to consider other distributions. So I described Dembski as having two methods, which I called the "chance elimination method" and the "uniform probability method". For the sake of completeness, I discussed and refuted both methods, not just the uniform probability one.

A common practice of those who have no good argument is to write ambiguously and equivocate between two (or more) positions and/or arguments. Those who wish to thoroughly refute the arguments are then forced to waste time sifting carefully through the writer's muddled text, trying to identify all the possible interpretations, and address each of them. It's particularly frustrating when a defender then points to one interpretation, calling it a straw man, and ignores the response to the other interpretation.

Ewert himself seems to have been confused by Dembski's equivocation, failing to observe that, according to Dembski's fully stated method, calculating a probablity under a given hypothesis only serves to eliminate that one hypothesis as an explanation for the observed event. We cannot draw a conclusion of "specified complexity" until we have eliminated all "relevant" hypotheses. As I wrote...

The conclusion of the Generic Chance Elimination Argument (step #8) is stated by Dembski as follows:

S [the subject making the inference] is warranted in inferring that E [the observed outcome] did not occur according to any of the chance hypotheses in {Hi}i in I and therefore that E exhibits specified complexity. [p. 73]

{Hi} is the set of all chance hypotheses which we believe "could have been operating to produce E" (p.72).

By Richard Wein (not verified) on 07 Jan 2016 #permalink

P.S. Given the vagueness and ambiguity with which Dembski describes his methodology, it helps to concentrate specifically on how he used his method to infer design in biology (in "No Free Lunch"). This consisted of two parts:
1. He attempted to put an upper bound on the probability of a bacterial flagellum arising by purely random combination of amino acids, i.e. he used a uniform probability distribution.
2. He made a non-mathematical argument from irreducible complexity against a flagellum evolving by natural selection.

Part #1 is uninteresting, since everyone agrees that purely random combination is an inadequate explanation for the origin of a flagellum. So all the work is really being done by part #2, the argument from irreducible complexity. Dembski's mathematical mumbo jumbo about probability, "specified complexity" and "information" turns out to be irrelevant to his main argument, which is just a variant on Behe's argument from irreducible complexity.

True, Dembski made a second argument, based on the No Free Lunch Theorems. But that's a quite different argument that didn't involve any probability calculations, and I refuted it separately.

It's also worth noting that Dembski seems to have largely given up on making these old arguments, and has now switched over to some new mathematical mumbo jumbo about "active information".

By Richard Wein (not verified) on 07 Jan 2016 #permalink

just listening from the back. This is interesting in a he-said-she-said sort of way...

sean s.

By sean samis (not verified) on 07 Jan 2016 #permalink

Let’s try to end on a substantive note.

That's the problem in a nutshell, isn't it? We'd like a substantive reproducible heuristic for these definitions (SC, CSI, active information), so researchers can independently and objectively determine what counts as them vs. not them. But after 20 years of ID research, there still isn't any.

Dr. Ewart, if you review Dr. Rosenhouse's blog posts, here's my request. Instead of writing articles addressing false straw men or correcting misconceptions, how about you just tell us how to determine whether a string has SCI, SC, active information, etc. or not. Below is a 1,000-letter potential/hypothetical genetic string. Show me how to evaluate it using intelligent design concepts. Does it have CSI? How do I calculate that? Does it have SC? How do I determine that? If Does it have active information? How do I determine that?


I eagerly await the how of inferring design given a potential genetic string. Well, perhaps "await" is the wrong word as I'm not really expecting a substantive answer...

Hmmm...that somehow got truncated. Oh well, I'll leave it for now and if I get a substantive reply I will reproduce the entire thousand letters.

Trying to summarize this for myself: the way I understand it, “specified complexity” is a conclusion; it cannot “do any work” until that’s confirmed. From what I gather, much of the conversation and disagreement about “specified complexity” turns on how to confirm it. That’s a valid topic, but until it’s resolved, “specified complexity” is pretty much useless.

It seems pretty clear that there is no resolution in site, and much changing of topics in the hope of obscuring that fact.

sean s.

By sean samis (not verified) on 07 Jan 2016 #permalink

Eric: "We’d like a substantive reproducible heuristic for these definitions (SC, CSI, active information)..."

I would say, "If you were engaged in a genuine scientific endeavour you would have done such-and-such", not "we want you to do such-and-such". The latter implies that they might have something worth saying, if only they would refine their work. But they have nothing of merit, just vacuous nonsense. What I want is for them to drop the whole thing.

By Richard Wein (not verified) on 07 Jan 2016 #permalink

** in sight ""

By sean samis (not verified) on 07 Jan 2016 #permalink

Sean: 'That’s a valid topic, but until it’s resolved, “specified complexity” is pretty much useless.'

It won't be resolved, because equivocation over terms like "specified complexity" is the means by which IDers create the false appearance (to the unwary and to themselves) of having something worth saying.

By Richard Wein (not verified) on 07 Jan 2016 #permalink

Richard: “It won’t be resolved, because ...

That’s probably true, but it appears I have correctly stated what it is that creationists “engaged in a genuine scientific endeavour” would have to do. Until then, arguments based on “specified complexity” are useless.

sean s.

By sean samis (not verified) on 07 Jan 2016 #permalink

Richard Wein @7: I agree in your characterization of the situation. I just generally prefer forward-looking 'invitational' responses because they provide fewer excuses for the respondent to digress.

Would I bet money that an ID proponent is going to show up here and provide a reproducible methodology for analyzing that string for CSI or IC? Well, maybe not. But if they do show up, I'll certainly consider the methodology they present.

too funny! You might want to actually get out more - then you would couch your replies with words like "probably true" and even imagine "creationists engaged in science".

By Michael Fugate (not verified) on 07 Jan 2016 #permalink

Eric: “But if they do show up, I’ll certainly consider the methodology they present.

Agreed; we all should.

sean s.

By sean samis (not verified) on 07 Jan 2016 #permalink

Another problem, quite aside from assumptions of the correct distribution, is that IDers commonly misunderstand what would constitute a hit in the search space. Taking cytochrome c as an example, they usually calculate the probability of getting a protein with the human cytochrome c sequence at random, i.e. (1/20)^L, where L is the length of the protein. Some of the more sophisticated allow for varying sequences, but only, at most, those sequences actually observed in some organism. We of course have no way to calculate the number of possible but unrealized cytochrome c sequences. But even this is not the proper universe. We should also consider radically different proteins that would perform the same function as cytochrome c, and again we have no way to estimate that. But even that isn't the proper universe. We really should consider all sequences, of any length, that would have some useful biological function. That, at least, we have some way of trying to estimate by assaying random sequences for a variety of functions. Of course it turns out when you do this that you end up with an appreciable fraction that display a selectable degree of one or more of the tested functions.

By John Harshman (not verified) on 07 Jan 2016 #permalink

Much like assumptions about correct distributions, assumptions about the “search space” appear to be motivated by a desire to constrain or eliminate unwanted results.

sean s.

By sean samis (not verified) on 07 Jan 2016 #permalink

These arguments are of course exercises in mental masturbation. Sometimes, we get though. For example, Michael Behe's son is an atheist due to Dawkin's book. So the fight is not hopeless.

However, I recently listened to Ben Carson's wife who is obviously a very intelligent articulate woman. I am always flabbergasted at how an intelligent person who went to Yale....yada yada...can be a fundamentalist Christian. That in the end is the real issue.

The deepest intention of these ID efforts is to sustain an intellectually unsustainable belief system. So one engages in a baroque display of intellectual obfuscation.

I think, in the end, these people are consumers of knowledge who have no grasp of how knowledge is produced. It is techne vs episteme to use Aristotle's terms. A car mechanic or a brain surgeon has absorbed a practical knowledge of how to do something. They have no concept of how knowledge is obtained or justified and therefore cannot consistently apply that to their other idiosyncratically acquired beliefs - our birth into a family with beliefs is do we leave Plato's cave?

Techne's have no methodological means of critiquing their acquired beliefs.

We can only keep beating the drums.

Eric, I believe we know it has "specified complexity" if it (your hypothetical gene) produces a functional, complex (which seem to be very rare) protein. That seems specific to me.

By Michael K. (not verified) on 11 Jan 2016 #permalink