The Fruits of Flow (I.D. XXIII)
In Is Fitness a "Specification"? (I.D. XXII), I suggested that fitness, as in Darwin's "survival of the fittest," serves admirably as a specification in William A. Dembski's sense of that word in Intelligent Design. This comes as something of a surprise, since Dembski is an anti-Darwinist.
To Dembski, information that is both complex and specified reliably implies design, and thus an intelligence responsible for the design.
Information is that which, because it somehow gets selected from a range of possibilities, reduces our uncertainty by eliminating all the other possibilites.
Complexity obtains for an informational "event" when the range of alternative possibilities is suitably large, making the selection of the single possibility that gets actualized highly improbable. Think of a royal flush in poker, as against all the other possible hands.
Specification exists for an informational possibility when, "independently of the possibility's actualization, the possibility is identifiable by means of a pattern." The pattern is a specification if and only if some part of our background knowledge can generate it in the absence of any knowledge of the event itself.
But, I said, an adaptation event during the course of Darwinian biological evolution by natural selection possesses a Dembskiyan specification! Namely, a "target" pattern which Darwin termed "fitness" serves as adaptation's specification, by virtue of the fact that we can say, independently of the actual adaptation, what adaptive changes would improve a population's fitness.
But, in this case, the real-world "target" is an unwitting one. There is no consicous archer intentionally shooting arrows at a pre-existing, visible target. Rather, blind natural selection simply removes the arrows that don't happen to strike the target. The fittest genetic variants are "selected for," and the less fit are "selected against."
Accordingly, we can look at, say, an adaptation event which famously occurred in England during the coal age. When burning coal blackened the countryside, once light-in-color moths turned dark. This was not because coal dust blackened them, but because they adapted genetically to produce more pigment and thus blend in better with their darkened surroundings, so to avoid predation.
But when England stopped its coal pollution and the countryside went back to normal shades, the moths went light again, for the very same reason.
In Dembskiyan terms, the mothly adaptation could be called a case of "respecification" — a word I made up; Dembski doesn't use it. At some earlier time, the pattern the moths adhered to was a specification for light-coloredness. Then came coal, and a new pattern/specification emerged: dark-coloredness. Finally came the abandonment of coal, and the old pattern/specification was reinstated.
The moths were unwittingly being aimed by "the survival of the fittest" — blind natural selection — at a moving target. Which they successfully hit.
The lesson here is that respecification does not necessarily imply intelligent design.
From Stuart Kauffman's work on self-organization, as reported in At Home in the Universe, we can conclude that initial specification need not imply design, either.
Kauffman shows that life may have originated with "collectively catalytic" or "autocatalytic" sets of proteins, wherein the production, out of smaller "food" molecules, of each and every protein in the set is catalyzed by some other protein in the set. When that vaunted attribute of "catalytic closure" is the case, the addition of just a few other simple structures and functions can make the autocatalytic set into a self-reproducing protocell ... and life and evolvability begin!
But catalytic closure is a target that is exceedingly likely to be struck whenever the number of interacting proteins in a set climbs into the low-to-middle double digits. It doesn't much matter which of the millions of available protein "species" are involved. For reasons having to do with network theory, Kauffman shows that if there are a moderately large number of proteins, any proteins, autocatalysis is virtually guaranteed.
So catalytic closure of a protein set (call it A) whose size is, say, 30 proteins, would seem to constitute an independent, Dembski-style specification for the set. The protein set itself would be highly complex, since the probability of a set arising which comprises exactly those 30 proteins is quite low. Thus, this set would seem to constitute, per Dembski, "complex specified information," or CSI.
In Holism, Emergent Phenomena, and CSI (I.D. XVI), I think I showed that another candidate set of 30 proteins (call it B) which does not happen to attain catalytic closure contains no less, or no more, information that does this autocatalytic one. That is, the two are equally (im)probable, thus equally complex.
Now I'd like to extend that insight by noting that the major Dembskiyan difference between A and B is that A's signature pattern, i.e., that of catalytic closure, is a specification which we can confirm as such by virtue of the background knowledge that Kauffman gives us, while B seemingly has no pattern, and thus no specification. B is complex, unspecified information whose information content — i.e., its complexity or improbability — is precisely equal to A's.
We can therefore conclude that the initial specification which can turn a haphazard collection of 30 proteins (B, for instance) into a 30-protein autocatalytic set (A) merely by the chance substitution of one protein for another in the original set does not necessarily imply intelligent design.
Putting these two cases together, that of the English moths and that of the moderately large protein sets, we can conclude that neither respecifcation nor initial specification of complex informational events necessarily implies design. Thus we can conclude, against Dembski's primary claim, that the existence of complex specified information, or CSI, does not always betoken intelligent design.
Dembski might object that the two examples I have given involve "microevolution," in the case of the moths, and self-organization, for the proteins, and so they are both cases in which there is a flow of CSI, but not its original creation. OK, I agree. The respecification in the one case and the initial specification in the other don't actually create new information, much less new CSI.
But I would go on to say that the "creation" of CSI is obviously not the same thing as the emergence of a (new or revised) specification. Nor is it the same thing as the "creation" of (additional) complexity by virtue of, say, a protein set growing from 29 members to 30, since a 29-member set, even without catalytic closure, is, all by itself, manifestly low-probability enough to qualify as CSI. Once information crosses the threshold from insufficiently complex to sufficiently complex in Dembski's scheme, it doesn't become "more CSI-like" by becoming yet more complex.
I'd say that the attaching of a new specification to a formerly unspecified complex informational event — making its sundry parts into a complex system — does not "create" any information at all. Take, for example, the dropping of a keystone into an arch. Before the keystone's arrival, the other stones in the arch were held up by scaffolding. Afterward, the scaffolding can be pulled away, and the arch stands on its own.
Dropping a keystone into a waiting arch does not create new information. It does, however, turn complex unspecified information into complex specified information.
The arch pattern, it is true, is a target struck by artifice, not by law or chance. The arch is not the product of either natural selection or self-organization. But, like the color-changed moths or the earth's putative first autocatalytic set, it is the fruit of flow.
Some complex specified fruits of flow are designed, and some are not. The arch is, the first autocatalytic set isn't.
In fact, I'd go so far as to say that emergence is what happens when, as a fruit of self-organized information flow, a specification appears spontaneously, in the absence of direction or design, thereby turning complex unspecified information (a protein set lacking catalytic closure, for instance) into complex specified information (an autocatalytic set). No new information is created, it is quite true, but information that was not specified becomes information that indeed possesses a specification. And, if Kauffman is right, no direction or design is implicated.
Yes, "order for free" magically appears; no, there is, on net, no new information. This analysis works by virtue of what I said in Information, Order, and Entropy (I.D. XXI). The self-organized system is — has to be — an exporter of entropy, which is the opposite of order and information. It "eats information" in its environment, so to build up its internal order. But, as a result, its environment is a net loser of information. The amount of information the self-orgainizing entity gobbles up, in order to produce its own "spontaneous specification," equals the amount of information its surroundings lose.
The fact that "sponteneous specification" — self-organized emergence — is possible is one reason why Dembski's claim — that to prove design all you need is complex specified information — is not quite complete. You also need what he alludes to briefly, but not at great enough length: complexity that is irreducible.
Irreducible complexity is the cornerstone of the argument to intelligent design made by Dembski's fellow proponent Michael Behe — see The Handmaiden of Design (I.D. XIV). Behe (per Dembski) uses the example of an ordinary mousetrap, none of whose five parts can be deleted without loss of all functionality. I'd say the classic arch is likewise an irreducibly complex system, since removing any single stone leads to its collapse.
I'd say that to prove design you need irreducibly complex specified information. Not just CSI, but ICSI, as it were.
But that fact, ahem, reduces Dembski's argument to Behe's. Behe, who wrote Dembski's introduction, clearly agrees with Dembski that the information content of irreducibly complex systems must be specified, so Dembski's contribution to intelligent design theory is far from negligible.
At the same time, it would seem to me that any "system" which deserves the name would have to be specified. Specification would have to be a synonym for "systemness." An arch is a system and not just a pile of stones because of the independently knowable pattern/specification/target which it hits, no?
Which means that, when it comes to implying design, irreducible complexity is really the name of the game.
As Dembski points out, it is to be distinguished from cumulative complexity. Cumulative complexity can be built up by blind Darwinian processes like mutation-cum-natural-selection one slow step at a time. Each step has to be "selected for," which presumably happens when it, all by itself, boosts the fitness of its unwitting innovator. Or, at least, when it doesn't reduce its possessor's fitness and therefore ability to compete with other variants for survival.
But — or so Dembski and Behe claim — mutation-cum-natural-selection cannot produce irreducible complexity. I know of no reason to dispute that claim. Darwinism's staunchest proponents (such as Richard Dawkins) apparently agree with it, after all. Dawkins's claim in books like The Blind Watchmaker is that complex — yea, even specified — biological structures like the eye are examples of cumulative complexity, not of irreducible complexity.
Dawkins shows what he thinks is a valid Darwinian path from the first slightly light-sensitive patch of skin to a full-fledged eye, complete with lens and retina. Each mincing step along the way improves its bearer's fitness in some palpable way. His conclusion is that an eye is not in fact irreducibly complex. It is (he would surely agree) a fruit of flow. What's more, it is a fruit of blind, undirected, undesigned flow.
To Dembski, information that is both complex and specified reliably implies design, and thus an intelligence responsible for the design.
Information is that which, because it somehow gets selected from a range of possibilities, reduces our uncertainty by eliminating all the other possibilites.
Complexity obtains for an informational "event" when the range of alternative possibilities is suitably large, making the selection of the single possibility that gets actualized highly improbable. Think of a royal flush in poker, as against all the other possible hands.
Specification exists for an informational possibility when, "independently of the possibility's actualization, the possibility is identifiable by means of a pattern." The pattern is a specification if and only if some part of our background knowledge can generate it in the absence of any knowledge of the event itself.
But, I said, an adaptation event during the course of Darwinian biological evolution by natural selection possesses a Dembskiyan specification! Namely, a "target" pattern which Darwin termed "fitness" serves as adaptation's specification, by virtue of the fact that we can say, independently of the actual adaptation, what adaptive changes would improve a population's fitness.
But, in this case, the real-world "target" is an unwitting one. There is no consicous archer intentionally shooting arrows at a pre-existing, visible target. Rather, blind natural selection simply removes the arrows that don't happen to strike the target. The fittest genetic variants are "selected for," and the less fit are "selected against."
Accordingly, we can look at, say, an adaptation event which famously occurred in England during the coal age. When burning coal blackened the countryside, once light-in-color moths turned dark. This was not because coal dust blackened them, but because they adapted genetically to produce more pigment and thus blend in better with their darkened surroundings, so to avoid predation.
But when England stopped its coal pollution and the countryside went back to normal shades, the moths went light again, for the very same reason.
In Dembskiyan terms, the mothly adaptation could be called a case of "respecification" — a word I made up; Dembski doesn't use it. At some earlier time, the pattern the moths adhered to was a specification for light-coloredness. Then came coal, and a new pattern/specification emerged: dark-coloredness. Finally came the abandonment of coal, and the old pattern/specification was reinstated.
The moths were unwittingly being aimed by "the survival of the fittest" — blind natural selection — at a moving target. Which they successfully hit.
The lesson here is that respecification does not necessarily imply intelligent design.
From Stuart Kauffman's work on self-organization, as reported in At Home in the Universe, we can conclude that initial specification need not imply design, either.
Kauffman shows that life may have originated with "collectively catalytic" or "autocatalytic" sets of proteins, wherein the production, out of smaller "food" molecules, of each and every protein in the set is catalyzed by some other protein in the set. When that vaunted attribute of "catalytic closure" is the case, the addition of just a few other simple structures and functions can make the autocatalytic set into a self-reproducing protocell ... and life and evolvability begin!
But catalytic closure is a target that is exceedingly likely to be struck whenever the number of interacting proteins in a set climbs into the low-to-middle double digits. It doesn't much matter which of the millions of available protein "species" are involved. For reasons having to do with network theory, Kauffman shows that if there are a moderately large number of proteins, any proteins, autocatalysis is virtually guaranteed.
So catalytic closure of a protein set (call it A) whose size is, say, 30 proteins, would seem to constitute an independent, Dembski-style specification for the set. The protein set itself would be highly complex, since the probability of a set arising which comprises exactly those 30 proteins is quite low. Thus, this set would seem to constitute, per Dembski, "complex specified information," or CSI.
In Holism, Emergent Phenomena, and CSI (I.D. XVI), I think I showed that another candidate set of 30 proteins (call it B) which does not happen to attain catalytic closure contains no less, or no more, information that does this autocatalytic one. That is, the two are equally (im)probable, thus equally complex.
Now I'd like to extend that insight by noting that the major Dembskiyan difference between A and B is that A's signature pattern, i.e., that of catalytic closure, is a specification which we can confirm as such by virtue of the background knowledge that Kauffman gives us, while B seemingly has no pattern, and thus no specification. B is complex, unspecified information whose information content — i.e., its complexity or improbability — is precisely equal to A's.
We can therefore conclude that the initial specification which can turn a haphazard collection of 30 proteins (B, for instance) into a 30-protein autocatalytic set (A) merely by the chance substitution of one protein for another in the original set does not necessarily imply intelligent design.
Putting these two cases together, that of the English moths and that of the moderately large protein sets, we can conclude that neither respecifcation nor initial specification of complex informational events necessarily implies design. Thus we can conclude, against Dembski's primary claim, that the existence of complex specified information, or CSI, does not always betoken intelligent design.
Dembski might object that the two examples I have given involve "microevolution," in the case of the moths, and self-organization, for the proteins, and so they are both cases in which there is a flow of CSI, but not its original creation. OK, I agree. The respecification in the one case and the initial specification in the other don't actually create new information, much less new CSI.
But I would go on to say that the "creation" of CSI is obviously not the same thing as the emergence of a (new or revised) specification. Nor is it the same thing as the "creation" of (additional) complexity by virtue of, say, a protein set growing from 29 members to 30, since a 29-member set, even without catalytic closure, is, all by itself, manifestly low-probability enough to qualify as CSI. Once information crosses the threshold from insufficiently complex to sufficiently complex in Dembski's scheme, it doesn't become "more CSI-like" by becoming yet more complex.
I'd say that the attaching of a new specification to a formerly unspecified complex informational event — making its sundry parts into a complex system — does not "create" any information at all. Take, for example, the dropping of a keystone into an arch. Before the keystone's arrival, the other stones in the arch were held up by scaffolding. Afterward, the scaffolding can be pulled away, and the arch stands on its own.
Dropping a keystone into a waiting arch does not create new information. It does, however, turn complex unspecified information into complex specified information.
The arch pattern, it is true, is a target struck by artifice, not by law or chance. The arch is not the product of either natural selection or self-organization. But, like the color-changed moths or the earth's putative first autocatalytic set, it is the fruit of flow.
Some complex specified fruits of flow are designed, and some are not. The arch is, the first autocatalytic set isn't.
In fact, I'd go so far as to say that emergence is what happens when, as a fruit of self-organized information flow, a specification appears spontaneously, in the absence of direction or design, thereby turning complex unspecified information (a protein set lacking catalytic closure, for instance) into complex specified information (an autocatalytic set). No new information is created, it is quite true, but information that was not specified becomes information that indeed possesses a specification. And, if Kauffman is right, no direction or design is implicated.
Yes, "order for free" magically appears; no, there is, on net, no new information. This analysis works by virtue of what I said in Information, Order, and Entropy (I.D. XXI). The self-organized system is — has to be — an exporter of entropy, which is the opposite of order and information. It "eats information" in its environment, so to build up its internal order. But, as a result, its environment is a net loser of information. The amount of information the self-orgainizing entity gobbles up, in order to produce its own "spontaneous specification," equals the amount of information its surroundings lose.
The fact that "sponteneous specification" — self-organized emergence — is possible is one reason why Dembski's claim — that to prove design all you need is complex specified information — is not quite complete. You also need what he alludes to briefly, but not at great enough length: complexity that is irreducible.
Irreducible complexity is the cornerstone of the argument to intelligent design made by Dembski's fellow proponent Michael Behe — see The Handmaiden of Design (I.D. XIV). Behe (per Dembski) uses the example of an ordinary mousetrap, none of whose five parts can be deleted without loss of all functionality. I'd say the classic arch is likewise an irreducibly complex system, since removing any single stone leads to its collapse.
I'd say that to prove design you need irreducibly complex specified information. Not just CSI, but ICSI, as it were.
But that fact, ahem, reduces Dembski's argument to Behe's. Behe, who wrote Dembski's introduction, clearly agrees with Dembski that the information content of irreducibly complex systems must be specified, so Dembski's contribution to intelligent design theory is far from negligible.
At the same time, it would seem to me that any "system" which deserves the name would have to be specified. Specification would have to be a synonym for "systemness." An arch is a system and not just a pile of stones because of the independently knowable pattern/specification/target which it hits, no?
Which means that, when it comes to implying design, irreducible complexity is really the name of the game.
As Dembski points out, it is to be distinguished from cumulative complexity. Cumulative complexity can be built up by blind Darwinian processes like mutation-cum-natural-selection one slow step at a time. Each step has to be "selected for," which presumably happens when it, all by itself, boosts the fitness of its unwitting innovator. Or, at least, when it doesn't reduce its possessor's fitness and therefore ability to compete with other variants for survival.
But — or so Dembski and Behe claim — mutation-cum-natural-selection cannot produce irreducible complexity. I know of no reason to dispute that claim. Darwinism's staunchest proponents (such as Richard Dawkins) apparently agree with it, after all. Dawkins's claim in books like The Blind Watchmaker is that complex — yea, even specified — biological structures like the eye are examples of cumulative complexity, not of irreducible complexity.
Dawkins shows what he thinks is a valid Darwinian path from the first slightly light-sensitive patch of skin to a full-fledged eye, complete with lens and retina. Each mincing step along the way improves its bearer's fitness in some palpable way. His conclusion is that an eye is not in fact irreducibly complex. It is (he would surely agree) a fruit of flow. What's more, it is a fruit of blind, undirected, undesigned flow.
posted by eric at 9:31 AM
0 Comments:
Post a Comment
<< Home