Intelligence and Design.

My copy of No Free Lunch arrived a few days ago, and there are a couple of posts I want to make about it, but the first thing that struck me, reading the preface, and not for the first time, is how little Dembski (and other Intelligent Design proponents) seem to know about either Intelligence or Design.

As it happens, I have a relevant background in both.  I’m a cognitive scientist, and I came into cognitive science from a background in educational psychology, so I’ve always been interested in intelligence – how it works, how it is measured, what factors affect it, etc.  And, somewhat unusually for a cognitive scientist, I also have a training in design – I trained as an architect, a design training that is specifically focussed on “problem solving”, but I also applied that training to other “design” modalities, including composing music, and writing children’s books that attempted to explain something, both to commission, and therefore with a “design brief”.

And in both areas, what is abundantly clear, is that learning is critical.

When a child is struggling, cognitively, we say she is “learning disabled”, or is a “slow learner”.  When we design a building, or a piece of music, or a piece of writing, we embark on an iterative process in which our output feeds back as input into the process of critical appraisal and re-appraisal that informs sometimes radical, more often incremental, changes to our current creation.

In other words, “intelligent design” is a process  in which feedback from the environment, including our own output, iteratively serves as input into the design process.  Both intelligence in general, and design in particular, are learning processes.

But to read Dembski’s preface, you would not know it:

How a designer gets from thought to thing is, at least in broad strokes, straightforward: (1) A designer conceives a purpose.  (2) To accomplish that purpose, the designer forms a plan.  (3) To execute the plan, the designer specifies building materials and assembly instructions.  (4) Finally the designer or some surrogate applies the assembly instructions to the building materials.  What emerges is a designed object, and the designer is successful to the degree that the object fulfills the designer’s purpose.

Well, not exactly, IMO, and the part that Dembski misses (or, at best, glosses over) is precisely the part that most resembles evolution: the iterative feedback from the environment that results in the incremental adjustment of the prototype so that it ever more closely fulfils some function.  Not only that, but that function is not by any means always the original one.  For a building, typically it is, at least for its first occupants.  But buildings that survive the longest and are best maintained are those that are readily incrementally adapted for other functions.  And anyone who has ever made a pot, or carved a block of wood or marble, knows that what emerges is the result of a kind of dialogue between the sculptor and the material, and the result may be something very different to what the designer had in mind when she started.  Click on my sister’s blog in the blog roll if you don’t believe me 🙂

In fact, I’d go so far to say that the one thing that separates “intentional” design” from, I dunno, “iterative” or “tactile” design is that humans are capable of simulating the results of their iterative design before execution, so that we don’t have to build first, then dismantle.  But even then, we actually make models, often very crude models, out of crude materials, in the early processes of a design (well, this is true of architecture any way) – three dimensional back-of-the-envelope sketches, made of corrugated cardboard, bits of mesh, gauze, sponge, silver paper, prototypes we can nudge and fix and re-order and reassemble, according to how well the thing seems to work.

Intelligent design is very like evolutionary processes, in other words.  So it’s not surprising that the products of both should show a family resemblance.  Oddly, I agree with Dembski that he has put is finger on a kind of pattern that is distinctive, when he talks about “specified complexity”.  I just don’t think it has much to do with intention, and everything to do with iterative adjustments in response to environmental feedback.

Biology has all the hallmarks of a learning process, in other words.  Evolutionary processes are learning processes, as is human intelligence.  Those would seem to be reasonable candidate authors of a pattern that exhibited “specified complexity”.  An omniscient and omnipotent creator, not so much.

 

 

 

180 thoughts on “Intelligence and Design.

  1. William J. Murray: Insistence and repetition is not an explanation. Nobody has explained anything – what “what works best” means, or what “optimising” means – in non-normative, non-vague rigorous terms, not a collection of just-so stories.What is the rigorous meaning of those terms when it comes to NS?

    William, yet again, you are completely missing the point. Natural selection, by definition is the selection of what currently works best. Rigorously, that means that each generation of individuals in a population will possess a distribution of variants that is biased by the degree to which the parental bearers of those variants were successful in reproducing.

    That is what natural selection is. Joe is simply wrong when he says “natural selection doesn’t do anything” or that “good enough” is all that results from it. Clearly, the most successful breeders in one generation will leave a greater representation of their genomes in the next generation than those who are less successful. That is so obviously true that just like that, it is a tautology. However, natural selection is not a tautology because it refers to a bias in favour of genomes that tend to promote reproductive success in the current environment.

    So if we have a population of N individuals with a distribution of M genotypes, and all genotypes produce phenotypes with equal probability of successful reproduction, nonetheless the next generation will not have the same distribution of genotypes as the previous, simply because inevitably some genotypes will get luckier than others. But the sampling of genotypes in the new generation will be unbiased in favour of any one genotype.

    However, if all genotypes do NOT produce phenotypes of equal probability of successful reproduction, that next generation will have a biased sampling of genotypes, where the bias is in favour of those genotypes that confer greater probability of reproductive success on their phenotype.

    And that is what is termed “natural selection” – heritable variance in reproductive success in the current environment. Obvious it “does something” – by definition it “does something” – it’s the process that results in a generational sampling of genotypes that is biased in favour of what best promotes reproductive success.

    And it is cumulative – what is best in one generation may be worst in later generations.

    I’ll come back later for the rest of your post, but this really needs to be put straight.

  2. William, yet again, you are completely missing the point. Natural selection, by definition is the selection of what currently works best. Rigorously, that means that each generation of individuals in a population will possess a distribution of variants that is biased by the degree to which the parental bearers of those variants were successful in reproducing.

    I didn’t miss that point. It’s irrelevant. However you define the “here and now” bias of NS, unless that bias can be shown to, over the course of time, categorically favor the current diversity of life, it adds nothing to its explanation.

  3. When one says that NS produces X (fitness, progeny success, a stabilized population, “optimisation”, “what works best”,or whatever one wishes to symbolize as X), they haven’t shown how X is categorically and necessarily related to Y (which represents the kinds (categories) of biological diversity which exists today.

    Unless NS is biased in favor of the kind of categorical diversity we see today, it adds nothing to its explanation. For all we know, NS works against such categories of features, and what we see today are just flukes that fell through the cracks in the selection process because of unlikely combinations of events. So-called “higher” life forms would just be the “mistakes” that slipped by the overall NS bias, whereas single-celled organisms would be the NS-favored organism category.

  4. WJM: I think you think NS is given more credit within biology (rather than in popular conception) for the diversity of life than is actually the case. Diversity is the product of stochastic processes, both biased (NS) and unbiased (Mutation, Drift and Gene Flow), operating differentially across a reproductive divide. That diversity does not demand complexity, but certain contingencies have favoured it in a limited set of rather large organisms. Of course we obsess about complexity, because we and everything that interests us is complex. But if you look at an unbiased ‘tree of eukaryotes‘ (let’s not side-track as to whether it is Tree, Bush or some other metaphor), you will see that the vast majority of diversity is unicellular. If you’re looking for ‘us’, we are at 4 o’clock inside Opisthokonts, the whole of animal life buried inside that word ‘Metazoa’, with us in the ‘Deuterostomes’. Fungi are just above us, and the Plants at 1 o’clock, from Bryophytes to Pteridophytes. With few exceptions, the rest is unicellular. If we were to embed this tree into the broader ‘tree of Cells’, the whole of the eukaryotes would be similarly dwarfed by the mass of prokaryotes. Complex organisms have corralled a substantial amount of the global biomass into their bloated bodies – about half, it is estimated. But numerically, in terms both of species and total genomes, God appears to have an inordinate fondness for single-celled organisms.

    Natural Selection does not predict or explain this tree, nor any other. Why should it? This is the purview of Common Descent. Bifurcations in the tree – ‘nodes of increasing diversity’ – are caused by separation of populations, and, by and large, Natural Selection acts only within a population. Nonetheless, NS does explain why organisms are better adapted to their environments than a collection squirted out of a ‘DNA randomiser’ might be – it explains ‘fitness of form’, not ‘why this particular set of forms’. Members of a species are close competitors – being genetically close forces them to be ecologically close. Just as ecological competition between species has the inevitable result of elimination of one or the other (the competitive exclusion principle), so ecological competition between varieties of the same species has the same, inevitable result – elimination of one or the other, irrespective of any ‘edge’ one or the other might have. That is: with or without Natural Selection, a species will change if nonlethal mutations occur with any frequency. But with Natural Selection (greater average offspring numbers for one variety than the other), the resulting population will be more ‘tuned’ to the environment in which it has to survive.

    I think it vital to grasp the importance of sampling. Take any finite sampling process – a deck of cards, say. Draw a card at random, and either discard (death) or duplicate and return (birth). Keep doing this, and you WILL end up with a deck consisting solely of copies of the same card. All other ‘varieties’ have been eliminated from the ‘population’. Introduce a bias for or against a particular suit or face value, and you increase or decrease the chances that the surviving card will be of that type – being that type of card has greater or lesser survival value in that selective environment. It produces more or fewer ‘offspring’.

  5. WJM:

    So-called “higher” life forms would just be the “mistakes” that slipped by the overall NS bias, whereas single-celled organisms would be the NS-favored organism category.

    There is not an ‘overall NS bias’ other than increased local survival and reproduction. If a non-bacterial organism arises that can survive and reproduce in a space that would otherwise be occupied by bacteria – by eating them, for instance – then NS would be expected to favour it. Bacteria cannot eat. It would not require an intelligent designer to protect it from the ravages of the prokaryotes; its enhanced survival does that.

    There is an interesting conundrum about what a ‘species’ or a ‘population’ actually is when there is no sexual reproduction, but if we simply look at it as an ecological matter, there are clear aspects of biology that allow single-celled eukaryotes to survive in a world of bacteria, and why their multicellular descendants gained a similar ability to survive in a world of unicells. They hit upon unoccupied ecological niches, and diversified. They are neither mistakes nor successes, just survivors – adaptation (NS) forming part of this process.

  6. Allan Miller:
    Joe GYou are getting desperate! If direct observation of an event, rather than observing reality and determining its consequences, was a valid objection to the existence of a phenomenon, ID would certainly not have a leg to stand on. Which it doesn’t, of course, because you have neither direct observation nor any means to analyse the definitive fingerprints of your favoured mechanism, except in cases that we don’t need the ‘theory’ for, because WE did it.

    If you stochastically sample with replacement, with or without bias (NS), you will get fixation, unless something intervenes to stop it. In finite populations, stochastic sampling is inevitable, and the survivors the only available source of replacement. So what stops fixation?

    Changing environments stop fixation- the observed wobbling stabity prevents fixation- IOW reality prevents fixation- that is in a blind watchmaker’s world.

    However bottlenecks and design could allow for a mutation to become fixed.

  7. There is not an ‘overall NS bias’ other than increased local survival and reproduction.

    Thank you. I appreciate direct, honest answers. This means that NS is not contributory to the explanation of so-called “higher” life forms. That doesn’t mean ID is involved; it just means that NS doesn’t add anything to the explanation of the generation of so-called “higher” life-forms as a broad category, or any of their particular features.

  8. They are neither mistakes nor successes, just survivors – adaptation (NS) forming part of this process.

    Well, it may be part of the process, but it’s not a meaningful part of the explanation. Gravity and erosion may be non-random (governed by force or molecular dynamics) parts of a process that ends up with a particular feature of the world, but that doesn’t mean they contribute significantly to the explanation of that feature.

  9. One of my points here is that, when someone says that some higher-life-form feature cannot be explained by random or chance evolution, Darwinists often retort that evolution isn’t a random or chance process, because of NS, and that the claim of randomness is erroneous.

    They might as well be saying that evolution isn’t random because gravity is acting on all life-forms, and gravity isn’t a random force. I mean … so what?

    It is not an erroneous point in proper context. NS doesn’t bias descent with variation towards any of the categorical features present in higher life forms, so all of those structures are random Y manifestations generated by a system that is only concerned with X. That air-breathing lungs exist is a random fluke in terms of the actual NS X-metric (producing successful progeny, “fitness”, etc.) Functioning wings are a random evolutionary fluke. Zebra stripes, self-aware mind, hooves, gills .. all just random evolutionary flukes built upon flukes.

    Even as a broad category, NS doesn’t necessarily or even more likely produce such things at all. It probably can be better said that NS acts against such higher-life-form contrivances in favor of the single-cellular life. Most of the higher life-forms that have ever existed are now extinct, whereas single-celled organisms still – pretty much – dominate the planet. It seems obvious from evolutionary history that so-called “higher” life forms have a very difficult time surviving for very long as species.

    The broader point is this: if evolution doesn’t bias evolutionary outcomes towards the production of [something like a functioning wing, heart or nervous system] (category), then darwinistic evolution’s capacity to explain the generation of such things can be no better than descent with variation with no NS acting on it whatsoever. In fact, it probably reduces the odds that such features will be generated.

    That such features are “possible” outcomes of an NS-sorted system is not an explanation. Possibility does not equal a scientifically plausible explanation, much less a probable one.

  10. William J. Murray: Insistence and repetition is not an explanation. Nobody has explained anything – what “what works best” means, or what “optimising” means – in non-normative, non-vague rigorous terms, not a collection of just-so stories.What is the rigorous meaning of those terms when it comes to NS?

    Apparently, what neo-darwinists mean when they say that NS selects for “what works best”,is “whatever feature or set of features happened to have survived”. Which, once again, is an invalid tautology.

    Unless one can show in a rigorous way that NS categorically biases evolutionary outcomes in favor of the kind of biological diversity that actually exists, they haven’t added anything to the explanation of such diversity. All they have done is make it easy for some to believe in NS as an explanation by offering post-hoc, tautological just-so stories filled with vague, normative concepts such as “fittest”, “optimising”, “works best”, “in order to stabilise”.

    I appreciate Alan Miller’s honest answer to my question.At least he didn’t try to duck and weave and give cover to ideology when he said:

    I think any honest person not serving some ideology would have to agree; that 2 billion years could have as easily, or more easily, been 4 or 10.

    That depends on whether or not OOL is as much (or more) of an aberration than endosymbiosis and sex. But, OOL is granted for argument’s sake in my discussion here.

    As I said before, NS doesn’t add to the explanation of the diversity of life if that diversity cannot even be predicted as a categorical bias from the first replicating organisms. We first have the apparent fluke of OOL; then NS supposedly takes over. Then we have the flukes (or aberrations, as you say) of endosymbiosis and sex.

    I would entirely agree that NS alone, doesn’t account for lateral diversity. What it accounts for is longitudinaladaptation. For instance, in my exercise I demonstrated the power of NS to optimise of a population to its environment, but it did not result in diversity. Indeed, my final population was very much less diverse than my starting population.

    On the other hand, NS accounted for a great deal of longitudinal diversity, and the individuals of my final population looked nothing like any of the individuals of my starting populations.

    Lateral diversity is not therefore an inevitable result of NS, and, oddly, Darwin’s book “on the origin of species” is less about the origin of species than of the longitudinal adaptation of each lineage. However, in practice, NS biased towards diversity because in nature, unlike in my exercise, there are a great many factors (not just one) that affect survival and which interact with genotype, and the environment therefore offers a great many niches to which subpopulations may become optimised.

    Speciation – and thus lateral diversity – occurs when a single population splits into two subpopulations in which within-sub-population breeding is much more prevalent than between-sub-population breeding (note that this definition of speciation does not apply to cloning species, in which every new genotype starts a lineage, is, in effect, a new “species”) resulting in each sub-populations to adapt independently.

    So adaptation is essential for speciation, but speciation itself depends on the population dividing into non-interbreeding lineages.

    For all we know, any normative concept we apply post-hoc such as “optimisation”, “fitness”, or “what works best” means, in actual, physical, rigorous NS terms, to keep everything at the prokaryotic level. It is only because non-prokaryotic other things exist that we imagine – again, normatively – that such diversity somehow must have been “more fit” than what is obviously the most fit organism on the planet in any real terms.

    You are misunderstanding the concept of “fitness”. It isn’t a “normaltive” concept at all – it’s entirely relative. An organism is “fitter” than its peer if it possess a genotype more likely to lead to its having viable offspring. But if you are talking about cloning populations (e.g. unicellular eukaryotes and prokaryotes), then a eukaryote and its cousin prokaryote are not in competition, so talking about their relative fitness is irrelevant. As long as either is viable, it and its peers will form part of two adapting (evolving) populations. “Optimisation” refers to the process by which the best (optimal) self-replicators (i.e. the fittest of each generation of a population are most copiously represented in the next. The concept of relative fitness doesn’t apply to separate populations. Of cours separate populations could still be in competition for resources, indeed one could be a resource for the other! But now we are not talking about relative fitness but eco-systems and their stability.

    Darwinists fill in the blank with “just-so” stories about evolutionary “fitness” and “optimization” as if they are rigorous, positive terms, but what the really are are just post-hoc rationalizations that“makes sense” to them in terms of why wings and hooves and zebra stripes might have come to exist.

    You are mistaking speculation about what selective pressures might have resulted in which features for a theory about how selection operates to optimise a population.

    Darwinian evolution demonstrably and a priori results in optimisation. However, post hoc, it is not possible, except in rare circumstances, to track back to find out which variants at which stage produced what degree of enhanced fitness in what environment. Criticising Darwinian theory for not being able to account for the lineage of given feature is like criticising gravitational theory for not being tell us where the source of a river must be, given its mouth. And yet gravity, acting on water over a terrain can completely account for the course of the river. We can demonstrate that gravity is capable pulling rainwater from a mountainous valley to the sea and carving a riverbed as it does so; similarly we can demonstrate that natural selection (heritable variance in reproductive success) is capable of pulling an ancestral population of similar organisms occupying one niche to a large number of later populations occupying many niches.

    Do humans “work better” than bacteria?Are they more fecund, or more hardy?Can they survive in more environments? Do they use energy more efficiently? Are they easier to reproduce? Are they less prone to catastrophic failure?

    Probably not. I suspect our lineage will die out completely (because most lineages do) whereas the lineages of at least some modern bacteria will almost certainly be around long after we have gone. But the question is irrelevant to the issue of fitness because fitness, as I said, is not “normative” – it is relative, and only meaningful in relation to members of the same population.

    If NS doesn’t categorically predict a world full of the kind of life forms that exist, then those things are simply flukes. NS doesn’t explain their existence, even if it allows for it.

    Gravity doesn’t explain why this meteor created that crater. It simply explains why meteors create craters.

    That doesn’t make gravity an inadequate theory. Nor is Darwinian evolution.

  11. If you’re searching for X (progeny success) and you happen to, along the way, find Y (higher-life-form category of features), unless the search for X is also, in some way, a search for Y (biasing outcomes towards finding Y), then while the search for X is not random in terms of finding X, it is random in terms of finding Y.

    So, NS cannot be any better (and might be worse) than any random search (in terms of finding higher-life-form categories of features) unless it is somehow biased towards finding those features. Otherwise, that it is not a random search for “progeny success” is absolutely and totally irrelevant to the question of how higher-life-form features were generated.

  12. William J Murray,

    An analogy for NS can be “panning for gold”.

    The “environment” is a little pan with holes in the bottom, small enough for sand to fall through, but not large gold nuggets.

    The holes don’t know you are “searching” for gold, but they will not allow the sand to survive in the environment of the pan.

    Does that make sense to you?

  13. William J. Murray:
    One of my points here is that, when someone says that some higher-life-form feature cannot be explained by random or chance evolution, Darwinists often retort that evolution isn’t a random or chance process, because of NS, and that the claim of randomness is erroneous.

    Well, the problem here is the words “random” and “chance”. Both are misused, both by Darwin skeptics and Darwinists (Dembski in the first category IMO, Dawkins in the latter category). “Chance” is not an explanation even where we know what is meant (you yourself have defined it as “non-design”). Nor is “random”. Darwinian evolution is a “chance” process under the definition that “chance” means “non-design”, but that tells you nothing. Darwinian evolution is not a random process, if random means “equiprobable”. If random means “outcomes are drawn from a probability distribution” then it is, and the better (less ambiguous) word is “stochastic”. And both mutation and natural selection are stochastic processes.

    That means that both are drawn from probability distributions, and are both quite heavily constrained.

    Most new variants will confer fitness similar to the parental fitness. For a well-adapted population, most new variants will be neutral or slightly deterious, and only a very small proportion will be beneficial. For a not-yet-well-adapatd population, a larger proportion of new variants will be beneficial. For a population undergoing environmental change, some old neutral and even slightly deleterious variants will become beneficial and other old neutral and beneficial variants will become slightly (or greatly) deleterious.

    For Natural Selection (which is really inseparable from variance effects), the process is also stochastic, in that there is only a loose relationship between genotype and phenotypic fitness. A perfect specimen may be hit by a meteor. A runt may get lucky. But, averaged across many individuals, some variants will be consistently more successful than others, and so we can say that those variants are fitter.

    So in only once sense is NS “non-random” – unlike variance generation, NS is the word we use to refer to the biasing effect of variants that tend to promote reproductive success in the current environment. Mutations are also biased, but the bias is simply a function of the current adaptive status of the population – they will be biased against fitness in a well-adapted population, but possibly biased in favour of fitness (or less biased against) in a not-well-adapted population.

    They might as well be saying that evolution isn’t random because gravity is acting on all life-forms, and gravity isn’t a random force. I mean … so what?

    It is not an erroneous point in proper context. NS doesn’t bias descent with variation towards any of the categorical features present in higher life forms,

    No, that is correct. NS is simply a biased sampling of variants that tend to promote reproductive success in the current environment. That might include lungish things, but it might include something quite different. It might include efficient flow-through lungs, or, for a less lucky lineage, clunky old mammalian lungs.

    so all of those structures are random Y manifestations generated by a system that is only concerned with X. That air-breathing lungs exist is a random fluke in terms of the actual NS X-metric (producing successful progeny, “fitness”, etc.)

    Just as a particular river course is a “random fluke”. But that doesn’t mean that “no-river-course-at-all” is more likely. And if you are calling “fitness” an “NS metric” (I’ve been puzzled by what you have been meaning by NS as a metric), yes, if lungish things tend to promote reproductive success in a population, then the individuals with lungish things will become more prevalent in the populations. It may turn out that in most evolutionary systems (if we are not the only one in the universe) lungs are fairly common. Even in our own, eyes are fairly common, suggesting that gadgets that utilise reflected electromagnetic radiation in a frequence range that allows for high resolution to inform the critter about the distal environment are fairly common; ditto for gadgets that utilise ambient chemicals (sense of smell), and fluid-borne vibrations (sense of hearing). All seem readily generated and clearly useful for survival and thus successful reproduction.

    Functioning wings are a random evolutionary fluke. Zebra stripes, self-aware mind, hooves, gills .. all just random evolutionary flukes built upon flukes.

    Sure. Just as the precise course a river takes is a fluke – flukes on flukes. A boulder here, a porous stratum there, an outcrop somewhere else. The course a river takes is the result of a highly stochastic, but it nonetheless always runs down hill. Similarly, the course a population takes to survive a succession of environments is highly stochastic, but it nonetheless always runs in the direction of greater adaptation.

    Even as a broad category, NS doesn’t necessarily or even more likely produce such things at all.It probably can be better said that NS acts against such higher-life-form contrivances in favor of the single-cellular life.

    And that’s your mistake. You are confusing Natural Selection (a process) with fitness (a metric) and misunderstanding fitness. Fitness is measured relative to the other members of the population, not to members of other populations. That means that if a lineage of single-cell critters continues to adapt to its changing environment (or maintain its adaptation to a stable niche) it will not go extinct, and we will see its descendents today. It doesn’t mean that a lineage of multicellular critters will survive for millions of years if a) a multicellular variant comes about and b) that variant proves successful in its current environment and c) the population continues to adapt (and almost certainly subdivide and diverge) as new environmental opportunities arise.

    Most of the higher life-forms that have ever existed are now extinct, whereas single-celled organisms still – pretty much – dominate the planet. It seems obvious from evolutionary history that so-called “higher” life forms have a very difficult time surviving for very long as species.

    You are not comparing like with like. Single-celled organisms are largely asexual reproducers, and so the term “species” must be used with caution. Most lineages of single celled organisms must be extinct. The fact that some lineages have survived (and continue to diversify and adapt) doesn’t mean that all lineages have. Quite the reverse. The overwhelming majority have gone extinct.

    Most lineages of multicellular sexually-reproducing organisms, have also, as you say, gone extinct. However, many, as with unicellular lineages, are still going strong, and they dominate the planet in biomass at least. The largest proportion of the biomass is probably multicellular plants. Ants do pretty well too. So do we (I think we are slightly beaten by krill). But it depends what you mean by “dominate”. We are all part of a vast, interrelated ecosystem, and top predators (us, right now) are probably best described as the “dominators”. But that’s a quite different issue than “fitness”.

    The broader point is this: if evolution doesn’t bias evolutionary outcomes towards the production of [something like a functioning wing, heart or nervous system] (category), then darwinistic evolution’s capacity to explain the generation of such things can be no better than descent with variation with no NS acting on it whatsoever. In fact, it probably reduces the odds that such features will be generated.

    No. It does not reduce the odds. It greatly increases the odds of “such features” although possibly not of any one specific feature. Do you want me to run my exercise in the absence of NS? I can guarantee that the “feature” of runs-of-heads dominated by 4s and 3s will never emerge. And yet it hits the jackpot, reliably, with NS. And yet I cannot predict the order of 4s and 3s of the winning solution, because there are several solutions. So it is with biological features (and with evolutionary algorithms more complex and useful than mine) – there are often many solutions to the problem and a different one will emerge each time. What NS does vastly increase the odds that a solution will emerge. That might be lungs, wings, camouflage, long legs, short legs, big teeth, four stomachs, or whatever.

    That such features are “possible” outcomes of an NS-sorted system is not an explanation.Possibility does not equal a scientifically plausible explanation, much less a probable one.

    You are expecting Darwinian evolution to explain specific features. Mostly it doesn’t (although with modern genetics we can find out quite a lot). Darwinian theory explains why adaptive features emerge. Alone, doesn’t explain why one particular lineage should solve its environmental problem with legs, and another with flippers, although clearly it can explain why, in one environment, flippers will be more advantageous than legs and vice versa. As I said, post hoc, you can see why the river went the way it did. But gravity alone won’t tell you.

  14. William J. Murray:
    If you’re searching for X (progeny success) and you happen to, along the way, find Y (higher-life-form category of features), unless the search for X is also, in some way, a search for Y (biasing outcomes towards finding Y), then while the search for X is not random in terms of finding X, it is random in terms of finding Y.

    So, NS cannot be any better (and might be worse) than any random search (in terms of finding higher-life-form categories of features) unless it is somehow biased towards finding those features. Otherwise, that it is not a random search for “progeny success” is absolutely and totally irrelevant to the question of how higher-life-form features were generated.

    For a start, “search” is a bad metaphor. X is much more like an “attractor basin” than the object of a search. Populations are pulled towards X (progeny success) via variants that promote success in the current environment, much as the water in a river is pulled towards sea level by gravity, taking paths that it finds opportunistically (not always the shortest paths). Those paths may or may not include Y-like or proto-Y-like variants, but, if they do, that path is likely to be taken.

    And by “path”, I mean small steps in mutation terms i.e. between genomes that only differ the kind of respects that the reproduction-process tends to vary.

    Now, if no such paths exist, then no amount of pull from the attractor basins is going to help. The question is: why should we think that in biology, such paths do exist? There are two reasons, as I see it, to think so:

    One is that similar genotypes tend to produce similar offspring, and so a small mutation to a successful genotype will tend to produce a similarly succesful offspring. The second is that the genotypes are highly “connected” by various mutation types: insertion, deletion, substitution, duplication and recombination. In other words, if we plotted all possible genotypes (theoretically anyway) and connected each one to each other that was easily reachable by means those mutation types, we’d end up with a highly connected web. Vast areas of that web will be useless, but because of the first fact (that similar genotypes tend to produce similar offspring), successful genotypes (i.e. those that tend to promote progeny in the phenotype) will be clustered on that web. In other words, the “fitness landscape” is necessarily fairly smooth. Some parts may be detached and inaccessible (I’d rather like a flagellum myself, but I’m not going to get one, because wheels for multicellular organisms are probably IC) but as long as the simple corner of the web is viable, and connected to other viable parts of the web, populations starting at that simple corner (our earliest common ancestors) will be able to “travel” to many different parts of the web.

    What we don’t know, of course, is what that simple corner consisted of. But if we posit that there was a simple enough corner at some time (whether natural or divine) there is no reason that I can see to think that the rest of the web would not have been as accessible as it seems to have been.

  15. William J. Murray: Thank you.I appreciate direct, honest answers. This means that NS is not contributory to the explanation of so-called “higher” life forms. That doesn’t mean ID is involved; it just means that NS doesn’t add anything to the explanation of the generation of so-called “higher” life-forms as a broad category, or any of their particular features.

    No, it doesn’t mean that. Why should it? It just means that “higher life forms” were the route some lineages took as they were pulled along through various environments by your “X”.

  16. I wonder if WJM will ever answer any of the accumulating questions that have been directed at him.

  17. No. It does not reduce the odds. It greatly increases the odds of “such features” although possibly not of any one specific feature.

    Finally some text that actually addresses the point being debated.

    Do you want me to run my exercise in the absence of NS?

    Why? I don’t consider your “exercise” to be meaningful in this debate. If you want to conflate “accumulating strings of H” with “generating functional wings, hooves and circulatory systems”, and conflate a command that removes low-H strings from the field with “natural selection”, and then imagine your exercise has demonstrated something about evolutionary tendencies, go ahead. I don’t consider it meaningful here.

    What NS does vastly increase the odds that a solution will emerge. That might be lungs, wings, camouflage, long legs, short legs, big teeth, four stomachs, or whatever.

    Notice how you are conflating “solution to progeny problem” with “production of lungs, wings, camoflage”, etc., as if one necessarily had anything to do with the other. This is the core of your logical error.

    It greatly increases the odds of “such features” although possibly not of any one specific feature.

    NS can as easily select away from the category of features in question as towards it, and can as easily degrade anything it builds as continue building upon it.

  18. Notice how you are conflating “solution to progeny problem” with “production of lungs, wings, camoflage”, etc., as if one necessarily had anything to do with the other. This is the core of your logical error.

    Huh? The range of possible solutions to the progeny problem has nothing to do with the solution to the progeny problem? Are you sure this is what you meant here? How can solutions to a problem have “nothing to do” with solutions to that problem?

    NS can as easily select away from the category of features in question as towards it, and can as easily degrade anything it builds as continue building upon it.

    Yes, of course. Any given phenotypic change has as much chance of making an organism simpler, as of making it more complex. There is certainly no principle that complex always works better.

  19. William J. Murray:

    No. It does not reduce the odds. It greatly increases the odds of “such features” although possibly not of any one specific feature.

    Finally some text that actually addresses the point being debated.

    William, several of us have made this point several times. I’m pleased you now seem to be paying attention.

    *growl*

    Do you want me to run my exercise in the absence of NS?

    Why?

    Because you claimed that NS wouldn’t increase the odds of certain features evolve. If I leave out NS in my exercise, those features don’t evolve. If I include it, they do. Ergo, NS increases the odds.

    I don’t consider your “exercise” to be meaningful in this debate. If you want to conflate “accumulating strings of H” with “generating functional wings, hooves and circulatory systems”, and conflate a command that removes low-H strings from the field with “natural selection”, and then imagine your exercise has demonstrated something about evolutionary tendencies, go ahead. I don’t consider it meaningful here.

    That’s because you don’t understand evolutionary theory. If you did, you’d see the relevance. I am “conflating” nothing and my script does not “accumulat[e] strings of H”. It evolves a 500 character sequence in which the products of the runs of Hs is maximised. As a result, the evolving populations of sequences contain gradually more Hs in total; the runs of Hs start to focus on the 3s and 4s, and ultimately the 4s; the runs of tails between between runs of Tails minimises. There is no command in the script that tells the algorithm to find these features: the algorithm simply provides an environment in which the “product of runs of heads” is most likely to result in successful progeny.

    I know math and science aren’t your thing, William, but this is very simple math and very simple science, and you are smart.

    Noticehow you are conflating “solution to progeny problem” with “production of lungs, wings, camoflage”, etc., as if one necessarily had anything to do with the other. This is the core of your logical error.

    It isn’t an error at all. I am not “conflating” anything, as I said. Please read my post more carefully. There is no logical error.

    And in case it is still not clear to you, I’ll try to explain again, very specifically:

    The reason that “the solution to progeny problem” is necessarily related to the emergence of phenotypic features (whether “lungs, wings, camouflage” or my various sequences of runs-of-heads) is that phenotypic features are what affects the probability of success of the progeny. This is why it’s important to distinguish between: genotype; phenotype; fitness function; fitness landscape (and why Dawkins’ weasel is a poor example because genotype, phenotype, and fitness function are identical and the fitness landscape is perfectly smooth. This is grossly atypical).

    A genotype will tend to result in a phenotype that will tend to have a certain probability of reproductive success in the current environment. If, in the current environment, the probability of successful progeny is enhanced by phenotypic features that promote flight; camouflage; oxygen transfer from air; or high product-of-runs-of-heads, then genotypes that result in those phenotypic features will become more prevalent in the population.

    Moreover, if similar genotypes tend to result in similar phenotypes, and if similar genotypes are generated by the kinds of mutagenic processes involved in self-replication in that population, then the fitness landscape will tend to be fairly smooth, and evolution of phenotypic features that promote reproductive success will tend to evolve.

    There is no “logical error” here, and my toy example has all the features of a full model: genotype (the sequence of Hs and Ts); phenotype (the sequence of totals-of-runs-of-heads); an environment that favours reproduction of phenotypes in which the totals-of-runs-of-heads is large; and a range of mutagenic processes that mean that sequences with high totals-of-runs-of-heads are well-connected to sequences with lower totals, making the landscape smooth.

    I could have made my toy more complex by having a more complex environment, perhaps with several fitness criteria (for instance, in real environments you often find that a population will split to accomodate two kinds of foodstuffs, each best accessed by a slightly different phenotypic feature). It’s easy enough to do. But the principle is exactly the same. And anyone who has worked with real-life engineering applications of evolutionary algorithms (RBH for instance) will tell you how clever those populations can be at finding solutions to the problems their creators throw at them, including solutions that the creators did not intend – or want – them to find!

    NS can as easily select away from the category of features in question as towards it, and can as easily degrade anything it builds as continue building upon it.

    What “features in question?” Sure, looked at from the feature end, as it were, that can happen. Whale legs were “degraded” qua legs in order to produce a streamlined whale body. Sea-lion legs are also pretty “degraded” qua legs but form effective tail flippers. From a “leggist” standpoint, it’s a “degradation”, but from the population’s standpoint it’s an improvement, and the population’s standpoint is the one we are interested in. We wouldn’t call a whale with an atavistic pair of legs protruding ventrally a superior whale, any more than we’d call an antelope with whale-sized rear legs a superior antelope.

    If you are judging “NS” by its ability to evolve legs, then sure, it’s not going to reliably evolve legs if the population in question would be better off without them. But nobody claims that that’s what it does. What “NS” does is optimise the population to its environment by biasing the sampling in each generation of genotypic sequences that tend to produce the phenotypic features that maximise reproductive success in that environment.

  20. Underlying all this apparent confusion is the unspoken premise that evolution has goals, or that it is striving to achieve something. Or has targets.

    Get rid of the concept of target or direction and you can discuss what’s left over.

  21. Possibly guano-bound but I have to say that it has been a few years since i’ve seen anyone bury themselves beneath an avalanche of their own obfuscation as deeply as WJM seems determined to do.

    Natural selection is one of the easiest principles to grasp. It has been tested, observed and recorded in both the laboratory and in the wild. Unsurprisingly NS has been adopted in scientific and technological fields far and wide as a reliable method of optimisation and discovering novel solutions to difficult problems.

    But despite all this WJM’s vastly logical mind insists natural selection is some kind of phantom; an illusion; explanatorily bankrupt.

    It’s bizarre to watch.

  22. petrushka:
    Underlying all this apparent confusion is the unspoken premise that evolution has goals, or that it is striving to achieve something. Or has targets.

    Get rid of the concept of target or direction and you can discuss what’s left over.

    You (generic you) also need to get rid of the concept that you need a goal to get somewhere. A pull or a push will do just as well, and that’s exactly what natural selection is. It’s a tilt on the street that the drunkards are walking down. The drunkards don’t need a goal, they just need a slope.

    Or, conversely, as Homer Simpson found, there’s no escape from Fat Camp because all the exit roads run slightly uphill.

  23. Well, they don’t even need a slope to be more likely to end up a long way from where they started than nearby. But if we want to explain why they all end up at the south end of town, then we only need to posit a south-facing slope, not south-end pub.

  24. Not guano-bound. I’m not averse to general comments about a person’s arguments. That’s why I haven’t moved WJM’s comments on mine.

    Just growled.

  25. WJM

    One of my points here is that, when someone says that some higher-life-form feature cannot be explained by random or chance evolution, Darwinists often retort that evolution isn’t a random or chance process, because of NS, and that the claim of randomness is erroneous.

    Yes, as Elizabeth says, the word ‘random’ has so many connotations – everyone is convinced that their favourite of its 5 or 6 possible definitions is the one meant – that the charge is as easily simply accepted. Yes, evolution is random. So what?

    To a mathematician (the dominant skill of the evolutionary biologist), every probabilistic process, biased or not, is random. For less ambiguity, stochastic is preferred. It is in this sense that evolution is random. It is probabilistic, because even having a selective advantage does not secure the likeliest outcome every time. It is also random in the sense that there is no goal to the process – not even producing offspring. You will simply encounter the organisms whose ancestors did that, rather than those that did not. And mutation is random (not directed to a goal). And drift is random (it is sample error, a distortion unrelated to any qualities being selected for). As for NS … it is both random and nonrandom, according to preference. It does not cause just any old organisms to survive or die, but those possessed of particular genetically-based qualities. In causing the loss of lethal and severely disabling mutants, it is a preservative force, serving to keep a population where it is. In causing the promotion of more beneficial mutants (than the present incumbent) it serves to cause the population to ‘learn’ – to adapt, to preserve those features more useful in the current environment than those lost. Is that random or not? Depends on your definition, but ultimately … don’t care. It is what it is, to use a gratingly common piece of business-speak.

    And it is bound to have been involved in the transitions from simple prokaryote to simple eukaryote, and simple eukaryote to multicellular eukaryote – fluke transitions both, but substantial leaps in complexification – they created an opportunity for more complex life-forms, which do demonstrably gain advantage from their approach by exploitation of niches unavailable to bacteria. It is wrong to say ‘bacteria are the fittest organisms on earth’ – they can each produce at most TWO offspring, which is hardly spectacular.

    Natural selection did not cause the transitions, but the result of them was undoubtedly scrutinised by it, and the radical new way of life that each enabled and entailed tuned by it. And so it was with every ‘innovation’ since, arising initially in one species which then went on, in those very few instances, to found vast dynasties of descendant forms from these lucky (but useful) accidents. Wings are a fluke, lungs are a fluke, bilateral symmetry a fluke … every mutation and genetic recombination is an acknowledged fluke, so this is hardly a truth best left unsaid. Each fluke changes the game slightly. The error is in saying that “some higher-life-form feature cannot be explained by random or chance evolution”. So long as one does not equate ‘random’ with a mental image of a wrecking-ball producing a David from a block of marble (or a tornado in a junkyard!), it is entirely accurate to say that evolution is a series of flukes. NS stops it from going everywhere – it keeps things on viable tracks, and pushes them towards adaptation.

    In a great many cases, we can detect the precursors – it is not a total fluke, arising fully formed out of ash, but a retooling of that which existed already, to serve a different purpose. The signals of this co-option are splattered around the genome. Unicellular eukaryotic cells do not have actin so that we could have muscles – they use it in their own lives, for non-muscular purposes. Birds don’t have wings so that penguins could swim with them …

  26. Allan Miller:
    . In causing the promotion of more beneficial mutants (than the present incumbent) it serves to cause the population to ‘learn’ – to adapt, to preserve those features more useful in the current environment than those lost.

    Good, we are back to learning 🙂

    Yes, evolutionary processes form a learning system, very like our own – what is rewarded is repeated, what is punished, is made less likely to be repeated – at the level of synaptic connections as well as at the level of experience. This is one of the features of us as organisms enables us to act intelligently.

    We can also simulate, which enables us to conceive distal goals, and not merely react to proximal rewards and penalties. But the system is sufficiently similar that we use evolutionary algorithms as models of learning (also Bayesian algorithms, which operate on a closely related principle).

Leave a Reply