Richard Dawkins' views on non-random natural selection

[JimC]

mlpam wrote:

Second, natural selection's being a random process does not preclude it from producing adaptations; it just means adaptations are on average more likely than maladaptations. In fact, the fact that maladaptation is even possible is strong evidence that natural selection is a random process.

Thanks for confirming the non-random nature of selection!

[Psi Wavefunction]

mlpam wrote:

Second, natural selection's being a random process does not preclude it from producing adaptations; it just means adaptations are on average more likely than maladaptations. In fact, the fact that maladaptation is even possible is strong evidence that natural selection is a random process.

Thanks for confirming the non-random nature of selection!

... eh? Way to ignore the rest of that paragraph...

[Ilovelucy]

What part of "it's not a single mistake; it's that way that he has used it in several conversations in which he was elucidating his position" do you not understand?

Sigh.

What part of "single mistake" do you not understand? If Dawkins is mistaken about something, and says it twice, that mistake is still just a single mistake (one factual error).

Are you trolling?

WayOFTheDodo, if you think someone is trolling then use the report button. It is not for you to call people trolls on the boards or accuse them of trolling. In keeping with my earlier mod note, I have already binned earlier posts where you make this accusation for being off topic. Continuance of this will result in moderator action.

[Ilovelucy]

What part of "it's not a single mistake; it's that way that he has used it in several conversations in which he was elucidating his position" do you not understand?

Sigh.

What part of "single mistake" do you not understand? If Dawkins is mistaken about something, and says it twice, that mistake is still just a single mistake (one factual error).

Are you trolling?

WayOFTheDodo, if you think someone is trolling then use the report button. It is not for you to call people trolls on the boards or accuse them of trolling. In keeping with my earlier mod note, I have already binned earlier posts where you make this accusation for being off topic. Continuance of this will result in moderator action.

Additionally, as promised, I have once again binned all OT posts, and those replying to them. Among those posts were once again accusations of trolling, so what I just said to WOTD also goes for anyone else who wants to call others troll but neglects the report button.

[mizvekov]

mlpam wrote:

Second, natural selection's being a random process does not preclude it from producing adaptations; it just means adaptations are on average more likely than maladaptations. In fact, the fact that maladaptation is even possible is strong evidence that natural selection is a random process.

Thanks for confirming the non-random nature of selection!

You have just jumped at this thread without having read anything before, haven't you? Most of this thread has been dealing with exactly that question, and I though that it would be behind us by now, seeing as we are already at page 15.

Anyway, that definition of random you are using is equiprobable, and it's not what a scientist uses/should use.
In fact, it's not even the definition RD uses in most of this thread, which makes his responses inconsistent, and you have to guess which one he is using by having prior knowledge about what he is talking about.

[mizvekov]

Second, natural selection's being a random process does not preclude it from producing adaptations; it just means adaptations are on average more likely than maladaptations. In fact, the fact that maladaptation is even possible is strong evidence that natural selection is a random process.

By the way, susu said many times that random process and stochastic process are not the same thing, despite random and stochastic being synonyms. By the definition he put forward, a random process would be one which has undefined states, while stochastic would not. I can not however verify this anywhere else. Granted my only source of information on this is wikipedia, http://en.wikipedia.org/wiki/Stochastic_process and http://en.wikipedia.org/wiki/Random_process redirect to the same place.

[susu.exp]

In all of this discussion, will the sneering attackers of RD come up with a beautifully worded description, fit to explain to non-specialists, as to how their random processes produce the suite of complex adaptations I see in the real-world biota around me every day?

I´m no poet. But some of the mathematics is quite beautiful, I think. The question to ask there is how you would get the suite of compley adaptations in a deterministic process. The thing is, we´ve got the modern synthesis, which reconciled Darwin with Mendel. And this body of work describes evolution as a whole and selection in particular as a random process. There are problems if you do not do this, that are insurmountable (for instance if you do not treat offpring numbers as random variables and the fitness of an organism as the expected value of their offspring number, you could only have whole numbers as fitness. Generally adaptations arise through the fixation of alleles with small fitness benefits - when you only allow whole number fitness, you restrict evolution to huge leaps, rather than small gradual steps. And it turns out that an allele making that kind of difference is almost impossible).
The view expoused by you here is at odds with evolutionary biology. It´s no less at odds with it than the views of Behe (just as clueless about the mathematics underlying evolutionary theory).

"Nothing in biology makes sense except in the light of natural selection..."

So, neutral changes don´t exist? That´d imply that all molecular phylogenies are wrong and the large amount of agreement between the Chimp genome and the human genome would not indicate relatedness. It also implies molecular clocks don´t work. You´ve just mounted an attack on some of the best evidence for common descent out of ignorance.

And that´s the whole point of these "sneering attacks": Richard thinks the evidence for evolution is boring and it only interests some molecular biologists and mathematicians. Which in turn means he does a bad job at presenting the evidence (and spends his pages on things that are not directly evidence for evolution, but interest him). This lack of interest means that he gives some things the short end of the stick, even though he knows they exist (he does acknowledge neutrality in the molecular clock section for instance). This includes claims they don´t (which makes Chapter 2 of his book so annoying). What lay readers take from this is the impression that these things don´t exist and they reject the foundation of the molecular evidence for common descent. Which is what you just did. And if a book moves readers from being ignorant about the evidence for evolution, to being more ignorant about and it rejecting the evidence to boot - then this book deserves to be criticized.

You are an example for the problems that arise there. It´s throwaway remarks, but they cut to the core of evolution. And people reading that book come out rejecting evolution as it is understood in biology and embracing some odd alternative evolutionary theory, which simply doesn´t hold water and isn´t supported by any data.

There are flat earthers. If a book gave the evidence for the earth being a sphere, we could nitpick (it´s more of a flattened elipsoid, and actually it´s the geoid). But Richards book is in parts akin to claiming earth was a cube. And the reply to critics of the "cube earth" model is that it emboldens the flat earthers. But "cube earth" is not that much better than "flat earth"...

And there I stand...

...and there you´ll fall.

By the way, susu said many times that random process and stochastic process are not the same thing, despite random and stochastic being synonyms. By the definition he put forward, a random process would be one which has undefined states, while stochastic would not. I can not however verify this anywhere else.

I was mistaken there in the genera case. I did come across a specific paper which made this particular distinction and defined it´s terms like this upfront, but it´s not a common way to define them. I.e. I retract my earlier statement - they are generally synonymous and a differing use would require a specific upfront definition.

[natselrox]

@susu.exp:

I think I owe you an apology for the wrong statements I made. I wrongly proclaimed that the randomness of RND(0,1) is more than that RND(0.5,1). mizvekov made it quite clear to me. I am sorry.

I know, excuses are not welcome but I have been out of touch with maths for more than 2 years now! Medicine SUCKS!!!!!

Apologies to everyone. My statements were WRONG.

[mizvekov]

@susu.exp:

I think I owe you an apology for the wrong statements I made. I wrongly proclaimed that the randomness of RND(0,1) is more than that RND(0.5,1). mizvekov made it quite clear to me. I am sorry.

I know, excuses are not welcome but I have been out of touch with maths for more than 2 years now! Medicine SUCKS!!!!!

Apologies to everyone. My statements were WRONG.

Chill out, it could be true, depending on circumstances. I just showed that for normal floating point numbers, and that specific intervals, it would not hold.
We can just assume that you agreed if you did not object to the counter point.

[mizvekov]

@susu.exp:

I think I owe you an apology for the wrong statements I made. I wrongly proclaimed that the randomness of RND(0,1) is more than that RND(0.5,1). mizvekov made it quite clear to me. I am sorry.

I know, excuses are not welcome but I have been out of touch with maths for more than 2 years now! Medicine SUCKS!!!!!

Apologies to everyone. My statements were WRONG.

Actually, it's me who should apologize. After revisiting this, I realized that, because I was lazy and tried to calculate it in my head, I did miss that the interval [0,1] needs one bit of freedom on the exponent bias. So yes, assuming both intervals use the same machine representation, that interval has more states than the other.
But the point that I made earlier, and that susu also added to, still holds. Depending on what machine you use to do those calculations, it could leave the evaluation of the expression for the last step, applying optimization techniques which reduce the error. And who knows how it could be treating numbers and implementing that RND function.

Edit: Funny, but slightly on topic:

[natselrox]

@susu.exp:

I think I owe you an apology for the wrong statements I made. I wrongly proclaimed that the randomness of RND(0,1) is more than that RND(0.5,1). mizvekov made it quite clear to me. I am sorry.

I know, excuses are not welcome but I have been out of touch with maths for more than 2 years now! Medicine SUCKS!!!!!

Apologies to everyone. My statements were WRONG.

Actually, it's me who should apologize. After revisiting this, I realized that, because I was lazy and tried to calculate it in my head, I did miss that the interval [0,1] needs one bit of freedom on the exponent bias. So yes, assuming both intervals use the same machine representation, that interval has more states than the other.
But the point that I made earlier, and that susu also added to, still holds. Depending on what machine you use to do those calculations, it could leave the evaluation of the expression for the last step, applying optimization techniques which reduce the error. And who knows how it could be treating numbers and implementing that RND function.

Now I'm feeling dizzy!! It's midnight in India. So tomorrow..Have a great day!

[mizvekov]

Ok, just thought of something close to a formal proof for this problem of the randomness interval. I will try to keep it simple.

Generalizing for all numerical representation systems, the only thing that must hold true for all of them is that each state must represent at most one number.
One number can be represented by more than one state by the way. For example, zero is represented by two states in systems which use a 'sign' bit.
And ofcourse, a state might not represent any numbers, it could well be undefined (which would be wasteful), or represent something that might be interesting in the context of it's use, like +-infinity, NaN or other kinds of exceptions.

So from that, we can have any arbitrary mapping of states -> number.
One very odd example:

'0100' -> -7
'0101' -> 234.73
'0110' -> 150023.01010134

So, from here we can see that a very obvious counter example would be if the system did not represent any numbers in either intervals (In your example, [1,0])
Another counter example would be if the system represented only numbers which are present in both intervals (In your example, [0.5,1])

Apart from those two above pathological cases, in your example, it has to have more numbers in the interval [1,0], because it strictly contains the interval [0.5,1]. So you actually came very close here
Now if you had given the intervals [1,2] and [0,0.5], I leave as an exercise to you to conclude that you could absolutely not tell which one has more states, without knowing the state -> number mapping being used

[mjpam]

mlpam wrote:

Second, natural selection's being a random process does not preclude it from producing adaptations; it just means adaptations are on average more likely than maladaptations. In fact, the fact that maladaptation is even possible is strong evidence that natural selection is a random process.

Thanks for confirming the non-random nature of selection!

"RANDOM"=/="UNBIASED"

Unbiased systems and biased systems cannot be distinguished from one another a priori. For instance, if you draw with replacement from population of 10 individuals comprised of 5 A's and 5 B's and obtain a sample of 10 individuals comprised of 7 A's and 3 B's, you cannot tell a priori whether the composition of the sample is due to the way in which you sampled the population or the fact that there really is a bias toward drawing A's. In order to do so, you must examine the case in which there is no bias (i.e., the null hypothesis of neutral drift). In this case the probability of drawing at least 7 A's when there is no bias toward drawing A's is .1719, or approximately 1 out of 6 times. Another way of stating this is that probability of rejecting the null hypothesis when the null hypothesis is false is about .3836, or approximately 2 out of every 5 times. In other words, while you are about twice as likely to reject the null hypothesis when it is false as you to reject the null hypothesis when it is true, the fact that there is any probability of rejecting the null hypothesis when it is true is what makes natural selection a random process.

[seals]

@mjpam
Meaning 2 does not mean unbiased, susu already made the following point about it:

Oh dear - it looks like we were actually an entire definitional step further away from communicating than we realised! Meaning 2 isn't what 'statistically-minded' member of the thread were using - it's what they *thought* you might be using. Certainly, "random" is often used as a lazy shorthand for "uniform random".

Well, actually it is and re-reading some Fisher clarified this for me:
2. (statistics) Governed by or involving equal chances for each of the actual or hypothetical members of a population.
The hypothetical members of a population have associalted values, which are distributed through the cumulative probability function. In other words: The definition holds for non-uniform variables as well - and thus is equivalent to Meaning 3. It´s not how random variables are defined under Komolgorov, but they were defined in this ways before probability theory was axiomized.

But it does not matter really, because as you pointed out, Dawkins is inconsistently using the word random across the board, picking a different meaning almost every post.

Perhaps for clarification someone could list those 'different meanings' with their 'correct name' (ie not 'random') plus the one meaning that is correctly called 'random'?

It would be helpful if someone would then explain why Dawkins is 'flat out wrong' (assuming that's really the situation and this isn't afterall just pedantic nitpicking at some necessary adjustments to make the subject comprehensible to the non-expert ) in the same non-statistician's/layperson's language he uses. Until that happens there is the paradoxical impression that the erroneous explanation of how certain processes interact in evolution is so much more lucid and persuasive than the correct, horribly muddled version.

[mjpam]

It would be helpful if someone would then explain why Dawkins is 'flat out wrong' (assuming that's really the situation and this isn't afterall just pedantic nitpicking at some necessary adjustments to make the subject comprehensible to the non-expert ) in the same non-statistician's/layperson's language he uses. Until that happens there is the paradoxical impression that the erroneous explanation of how certain processes interact in evolution is so much more lucid and persuasive than the correct, horribly muddled version.

Been there, done that.

I'm also curious as to why you think an argument that continually equivocate among at least six different definitions of "random" (i.e., Dawkins' argument) is "lucid and persuasive", whereas an argument that uses only one definition of "random" (i.e., the one put forward by me and susu.exp) is "horribly muddled". At least with our argument, you don't have to keep wondering which definition of "random" we are using.

[jimmypippa]

Perhaps for clarification someone could list those 'different meanings' with their 'correct name' (ie not 'random') plus the one meaning that is correctly called 'random'?

It would be helpful if someone would then explain why Dawkins is 'flat out wrong' (assuming that's really the situation and this isn't afterall just pedantic nitpicking at some necessary adjustments to make the subject comprehensible to the non-expert ) in the same non-statistician's/layperson's language he uses. Until that happens there is the paradoxical impression that the erroneous explanation of how certain processes interact in evolution is so much more lucid and persuasive than the correct, horribly muddled version.

I am an engineer/physicist not a biologist but for me there are several issues.

Saying something is "the very opposite of random", when it actually fulfills the *technical* and thus precise definition of random as used throughout science, including biological sciences is misleading at best.

Richard Dawkins is an acknowledged expert on evolution, and I am not, so I needed cast-iron confidence in my ability to understand how the system of evolution works and why the usage of random was incorrect in that. It probably helped that I have investigated using evolutionary algorithms at work, so am not a complete novice.

The statement that "natural selection is the opposite of random", would if using the technical definition of "opposite to random" mean that all beneficial traits that occur would always be selected for.

This is obviously far from how natural selection happens. Indeed any particular individual trait, even a beneficial one is usually unlikely to survive past the first generation, due to simple Malthusian reasoning - a mutation carrying a theoretical 10% reproductive advantage in a single cod-fry still has odds of hundreds of thousands to one against it making past the first generation. (Several hundred thousand offspring and a stable or decreasing population, means that fewer than one offspring per parent actually makes it to reproduction). On balance though some beneficial traits will survive, and precious few deleterious ones.

If an event is random it is not predetermined. The opposite of random would thus be predetermined. In other words a statement that "natural selection is the opposite of random" would imply that natural selection is thus predetermined. In other words in a given environment, given organisms would always evolve to the same (presumed) optimal "configurations".

A related bugbear of mine is the phrase: "evolved to". Organisms didn't evolve traits to do anything. The traits evolved because organisms carrying genes for these traits reproduced more successfully. I think statements like these make it easy for the Intelligent Design proponents to claim that "God drives evolution by setting the natural selection criteria, just as engineers use evolutionary algorithms with artificially defined selection criteria".

I think that the claim that evolution is the opposite of random is actually attacking a strawman attack with an overly simplistic claim that is misleading, when the answer is to attack the creationist's definition of random as uniformly distributed. Even they know about gambling and that not all scores with pairs of dice are equiprobable.


EDITED to fix typo, see mjpalms post below: edited also to fix quotes

[mjpam]

I think that the claim that elocution is the opposite of random is actually attacking a strawman attack with an overly simplistic claim that is misleading, when the answer is to attack the creationist's definition of random as uniformly distributed.

I certainly hope that your elocution possesses some sort of bias to it!

[mizvekov]

Perhaps for clarification someone could list those 'different meanings' with their 'correct name' (ie not 'random') plus the one meaning that is correctly called 'random'?

It would be helpful if someone would then explain why Dawkins is 'flat out wrong' (assuming that's really the situation and this isn't afterall just pedantic nitpicking at some necessary adjustments to make the subject comprehensible to the non-expert ) in the same non-statistician's/layperson's language he uses. Until that happens there is the paradoxical impression that the erroneous explanation of how certain processes interact in evolution is so much more lucid and persuasive than the correct, horribly muddled version.

mjpam made a list several posts ago, but I'll try to sum some of them here.

A simple definiton of random, as used by scientists and mathematicians: That which is not deterministic. In other words, that which there is a probability of having more than one outcome.

Dawkins started using this definition: Unbiased. In other words, which has an equal probability for all outcomes. Dice throws with fair die fit here.

Then he claimed that drift was random by his definition. Following that, it was pointed out to him that drift is biased, specifically that drift is biased towards fixing higher frequency alleles.

Then Dawkins switched definition of random.
He started using random to mean: Statistically independent. This term needs another object to which you ascribe independence from. In other words, the outcome does not depend on this object. For example, dice throws are statistically independent from the position of Jupiter.

So by this definition, he said that drift is statistically independent from fitness. So far so good.
But then he claimed that in regards to adaptation, drift is useless, and it might just as well not be there.
Which is flat-out wrong, as even Lenski's experiment shows the importance of drift to adaptation, in this case of the evolution of the citric acid metabolizing E. Coli lineage.

Then lastly he threw again a bunch of dictionary definitions of random, and claimed those were the ones he was using all along. These include directionless, aimless, purposeless, etc. All which in this context means unbiased. So we go back to square one.

[mjpam]

I am reposting this with some minor edits, because I think that it clearly explains the many different way in which Dawkins is using "random", and some people seemed to have missed it the first time:

As mjpam also pointed out, Dawkins is inconsistent with his own definition , as he said that drift is random, but drift is biased towards the highest frequency allele.

Dawkins said drift to fixation is random with respect to adaptive improvement. However, as far as I can see, this doesn't mean he's saying it therefore also has to be random with respect to the frequency of the allele, or whatever. Whether it's random or not depends on what the randomness is in respect of.

... Now, the point you can learn from this is that, whenever we use the word 'random', it carries an implicit 'with respect to something'. When we are talking about evolution, especially when we are arguing with creationists, which unfortunately I usually am, the 'with respect to' that concerns us is 'with respect to adaptive improvement'. This is the 'with respect to' that leads us to say mutation is random, drift to fixation is random, and natural selection is nonrandom. This 'with respect to' is important, because complex adaptations are the creationists' prize exhibits. This was most certainly the context of the quotations from The Blind Watchmaker and Climbing Mount Improbable that you said were 'flat-out wrong'. ...

This issue is, in addition to being incorrect that "random" necessarily means "statistically independent, Dr Dawkins is equivocating. He started out by that saying that evolution is non-random because it is biased:

No it is not flat out wrong, you patronizing little twerp. It is absolutely and totally flat out correct. Note that I didn't say evolution was the very opposite of random, I said natural selection was. It makes all the difference. There are, as I am well aware ("having more than a passing acquaintance with the modern synthesis, the neutral theory and the nearly neutral theory"), various important ways in which randomness enters into evolution, in addition to mutation. The most important does indeed follow from the neutral theory, which I have publicly supported in several of my books. Mutations can drift to fixation in a population for reasons other than natural selection, and that process could indeed properly be called random. But natural selection, in precisely that sense, is non-random.

(original wording restored; red, original emphasis; bold emphasis added)

You apparently don't understand what randomness means. 'A bias in the probability' of something is pretty much exactly what we mean by non-random. Throwing dice is proverbially a random process. If you throw a die a thousand times, you expect to get a series of random numbers. If a particular die was biased towards, say, even numbers, it would deliver a non-random series of numbers. If natural selection is a bias in the probability of reproduction with respect to phenotype, that is equivalent to saying it is non-random. Do you really seriously not understand that?

(emphasis added)

It is a basic tenet on neutral theory that drift can and does fix alleles (some of them deleterious) in populations. In fact this undergraduate-level (?) population genetics textbook does a fairly straight forward job of explaining how this happens. Thus, evolution (defined as change in allele frequencies in a population) will occur even if there is no natural selection. It may not be able to produce the same forms that we observe in evolution by natural selection, but it still does happen.

Yes, mjpam, that is exactly what I said:

Mutations can drift to fixation in a population for reasons other than natural selection, and that process could indeed properly be called random. But natural selection, in precisely that sense, is non-random.

But, how many more times do I have to say this, I was not talking about evolution in those sentences in The Blind Watchmaker and Climbing Mount Improbable, which you described as "flat out wrong". I was talking about natural selection. Natural selection, not evolution: natural selection, natural selection, natural selection. Natural selection is non-random. Evolution is in many cases, for example when a neutral mutation drifts to fixation, random.

Do you get it now? When a gene goes to fixation because of natural selection, it is non-random. When a gene goes to fixation because of drift, it is random. Evolution consists of both processes, one random, the other non-random. I explicitly stated that I was talking about one of them, natural selection, the non-random one, the one defined as a bias (yes, a bias, that is, non-random) in reproductive rates.

(various emphases added)

In fact, the post you cited is the exception to the rule that Dr Dawkins is arguing that "random" means "unbiased", because, in his last post on in this thread, he goes right back to arguing that "random" means "unbiased":

Words, as Twatsworth rightly says, often have more than one meaning, sometimes related meanings. Confusion, and even patronizing abuse, can result when somebody adopts one meaning and presumes that another person is using the same meaning. The Shorter Oxford Dictionary gives two definitions of the adjective 'random'. In this order:

1. Not sent or guided in a special direction; having no definite aim or purpose.

2. (statistics) Governed by or involving equal chances for each of the actual or hypothetical members of a population.

Meaning 2 is the one adopted by statistically-minded members of this Forum, not surprisingly since that is the technical definition used in their profession.

Meaning 1 is the one used and assumed by everybody except professional statisticians. It is the one I have consistently followed in all my books, and the one understood by the kinds of people I am trying to communicate with: the kinds of people who need to be convinced of the truth of evolution, or who need better comprehension of what evolution means.

The two halves of Meaning 1 are themselves open to confusion. Meaning 1b ('having no definite aim or purpose') is the meaning assumed by creationists, who therefore regard evolution by natural selection as random, because it has no definite aim or purpose (which they assume to mean intelligently designed aim or purpose). Meaning 1a ('not sent or guided in a special direction') is the meaning adopted by most biologists, who therefore regard natural selection, but not mutation or drift to fixation, as nonrandom because it sends or guides evolution in the direction of adaptive improvement. It has been a large part of my life's work to dispel the confusion between 1a and 1b. So engrossed was I in the battle between 1a and 1b, I was momentarily taken aback by a sudden outflanking manoeuvre from an unexpected source, namely Meaning 2 (which I was aware of but had largely ignored and even forgotten about).

After some reflection, I shall continue to use Meaning 1, and shall continue my efforts to disentangle the confusion between 1a and 1b. I might think about possible ways to clarify the side issue of the confusion with Meaning 2. I don't think it is unkind to say that the postings to this forum by partisans of Meaning 2 are not well-adapted to enlighten laypeople. I can't help wondering whether it would be wise even to attempt to explain Meaning 2 to laypeople, while the more important confusion between 1a and 1b remains the dominant barrier to general understanding of evolution.

In fact, he further exacerbates the equivocation by (re)introducing third, fourth, and fifth glosses of "random" as "unguided", "purposeless", and "directionless". The problem with glossing "random" as "directionless", as already has been mentioned several times by several posters, is that even drift has a direction (regardless of whether it is a preferred direction), because it will eventually fix an allele at every locus. Thus, what is "random" using Dawkins' numerically predominant gloss of "unbiased" is simultaneously "nonrandom" using Dawkins' stated (and apparently preferred) gloss of "directionless".

Does anyone else have a problem that Dawkins' explanations would therefore lack clarity?

[seals]

It would be helpful if someone would then explain why Dawkins is 'flat out wrong' (assuming that's really the situation and this isn't afterall just pedantic nitpicking at some necessary adjustments to make the subject comprehensible to the non-expert ) in the same non-statistician's/layperson's language he uses. Until that happens there is the paradoxical impression that the erroneous explanation of how certain processes interact in evolution is so much more lucid and persuasive than the correct, horribly muddled version.

Been there, done that.

I'm also curious as to why you think an argument that continually equivocate among at least six different definitions of "random" (i.e., Dawkins' argument) is "lucid and persuasive", whereas an argument that uses only one definition of "random" (i.e., the one put forward by me and susu.exp) is "horribly muddled". At least with our argument, you don't have to keep wondering which definition of "random" we are using.

Well now you mention it, I didn't understand your explanation in that post at all. In the comments between the quotes I either thought you must be typing something other than what you meant, as it didn't seem to apply to the quote it was adjacent to, was referring to some word or statement I couldn't see in the quote, or I simply failed to follow your train of thought. Making sections bold and/or red and/or huge doesn't make it any clearer... that could just be me though, sorry!

But anyway on second thoughts, maybe someone should write, without reference to Dawkins and as if his version did not exist, their own stand alone version of the subject? Otherwise it would be weighed down and further complicated by irrelevant comparisons...

[mizvekov]

Well now you mention it, I didn't understand your explanation in that post at all. In the comments between the quotes I either thought you must be typing something other than what you meant, as it didn't seem to apply to the quote it was adjacent to, was referring to some word or statement I couldn't see in the quote, or I simply failed to follow your train of thought. Making sections bold and/or red and/or huge doesn't make it any clearer... that could just be me though, sorry!

But anyway on second thoughts, maybe someone should write, without reference to Dawkins and as if his version did not exist, their own stand alone version of the subject? Otherwise it would be weighed down and further complicated by irrelevant comparisons...

There are people here who argue you should use the technical definition of the word random, and concentrate on explaining what this random actually means. So in effect explain that evolution is a stochastic process, and more specifically, that it is a markov process.
There are others who think you should stick to some popular definition that lay people can relate to.

I prefer the former, but I don't have a strong opinion in this regard. But if the second is opted, then by no means the discussion should take any turn that involves other statistical terms, it would only lead to great confusion.
Another problem with using the second is that it gives opportunity for creationists to quote-mine. Also it contributes to a growing trend in evolutionary biology where terms which are wrongly used by creationists gets abandoned. One example of this (thanks to susu) is "irreducible complexity", which is seldom used anymore. This ofcourse adds again to the quote-mining problem, where old literature will be a source of those terms.

Now, if someone wants to go the easy (second) path, then by all means do so, but at least be consistent.
Pick exactly what you mean by "random".
if biased, then also exactly what kind of bias.
If statistically independent, then also with respect to what.
And don't ever introduce statistical terms which will conflict with these.

[mjpam]

Well now you mention it, I didn't understand your explanation in that post at all. In the comments between the quotes I either thought you must be typing something other than what you meant, as it didn't seem to apply to the quote it was adjacent to, was referring to some word or statement I couldn't see in the quote, or I simply failed to follow your train of thought. Making sections bold and/or red and/or huge doesn't make it any clearer... that could just be me though, sorry!

The main point of that post was that Dawkins chooses to change his definition of "random" according how persuasive he think his argument will be, which makes his over all argument completely unintelligible. When he was challenged on defining "random" as "unbiased", he switched gears and claimed that "random" meant "statistically independent". When it was pointed out to him that random processes were not necessarily statistically independent, he switched back to defining "random" as "unbiased" and threw in a couple more definitions, "unguided", "purposeless", and "directionless".

The biggest problem with his argument is that he wants to say that fixation due to genetic drift is "random" while fixation due to natural selection is "nonrandom". In trying to do so he has decided that genetic drift is random because the reproduction of the a particular instance of an allele is not biased by the type of the alleles, or, alternatively, the reproduction of the a particular instance of an allele is not statistically dependent the type of the alleles. However, he is trying to describe the aggregate process of ultimate fixation after many generation, which is both biased toward the higher frequency allele and statistically dependent on the type in so far as type of higher frequency allele is more likely to fix.

But anyway on second thoughts, maybe someone should write, without reference to Dawkins and as if his version did not exist, their own stand alone version of the subject? Otherwise it would be weighed down and further complicated by irrelevant comparisons...

The whole point of the thread, after Dlx2 said his peace about phylogenetics, was that Dawkins was misrepresenting evolution as a non-random process, so why wouldn't make reference to the things he has written and said?

[Dlx2]

As mjpam also pointed out, Dawkins is inconsistent with his own definition , as he said that drift is random, but drift is biased towards the highest frequency allele.

Dawkins said drift to fixation is random with respect to adaptive improvement. However, as far as I can see, this doesn't mean he's saying it therefore also has to be random with respect to the frequency of the allele, or whatever. Whether it's random or not depends on what the randomness is in respect of.

... Now, the point you can learn from this is that, whenever we use the word 'random', it carries an implicit 'with respect to something'. When we are talking about evolution, especially when we are arguing with creationists, which unfortunately I usually am, the 'with respect to' that concerns us is 'with respect to adaptive improvement'. This is the 'with respect to' that leads us to say mutation is random, drift to fixation is random, and natural selection is nonrandom. This 'with respect to' is important, because complex adaptations are the creationists' prize exhibits. This was most certainly the context of the quotations from The Blind Watchmaker and Climbing Mount Improbable that you said were 'flat-out wrong'. ...

That's not even "random" in the mathematical sense of the word so much as "noncorrelative." Correlation itself does not even directly imply mechanism; you have to go through a series of nontrivial inferential steps to get from point A (no correlation) to point Z (no mechanism). At the same time, correlation also does not mean causation or even nonrandom mechanism. There are a lot of ways to create a correlation with a significant p-value even when the correlation is being drawn on the basis of a few outlier and/or properties of the starting distribution.

Something to think about.

[sprite]

A few questions/points.

1. Could someone point to similar alternative books (ie aimed at the general public)to those of Dawkins which more 'correctly' explain evolution by using the 'correct' definition of random etc?

2. This argument has been reminding me of Dawkins' chapter in The Extended Phenotype - 'An Agony in Five Fits' where he discusses how he has not been using the term 'fitness' because of the confusion around the term and its many meanings.
To me this is something similar in that understanding a particular term often depends on the context but we can also react to the use of a particular term when it has become 'loaded' for some reason and the various interpretations can become an issue in itself.


3.One specific question and my main one:

In various models there is often the assumption of 'random mating'. What is then usually pointed out is that in real populations this is an assumption that often does not in fact exist - ie mating is not random.

The 'random' here in 'random mating' means what?
From my interpretation of this thread it does not mean equiprobable. Or does it? I'm sure sometimes I've seen it expressed as the possibility that any egg fuses with any sperm?

If not equiprobable then it means it might be biased? But not predetermined?

So when it is stated for the study of a real population that the mating is likely to be 'non-random', non-random means what?
Predetermined?
Adaptive? Is all this really an argument about what is 'adaptive' and what is 'nonadaptive'?

Perhaps someone could explain the 'random' and 'non-random' in respect of mating to help (me at least ) get some clarification of the use of this term in modelling.

[Cephus0]

A few questions/points.

1. Could someone point to similar alternative books (ie aimed at the general public)to those of Dawkins which more 'correctly' explain evolution by using the 'correct' definition of random etc?

2. This argument has been reminding me of Dawkins' chapter in The Extended Phenotype - 'An Agony in Five Fits' where he discusses how he has not been using the term 'fitness' because of the confusion around the term and its many meanings.
To me this is something similar in that understanding a particular term often depends on the context but we can also react to the use of a particular term when it has become 'loaded' for some reason and the various interpretations can become an issue in itself.


3.One specific question and my main one:

In various models there is often the assumption of 'random mating'. What is then usually pointed out is that in real populations this is an assumption that often does not in fact exist - ie mating is not random.

The 'random' here in 'random mating' means what?
From my interpretation of this thread it does not mean equiprobable. Or does it? I'm sure sometimes I've seen it expressed as the possibility that any egg fuses with any sperm?

If not equiprobable then it means it might be biased? But not predetermined?

So when it is stated for the study of a real population that the mating is likely to be 'non-random', non-random means what?
Predetermined?
Adaptive? Is all this really an argument about what is 'adaptive' and what is 'nonadaptive'?

Perhaps someone could explain the 'random' and 'non-random' in respect of mating to help (me at least ) get some clarification of the use of this term in modelling.

You can add me to the list of those interested in an answer to that problem. All very well to calculate the probability of an individual being autozygous in a population by random (with some trepidation in using the word) sampling of the effective population. Mating is rarely random though and what of deme size / interaction? Difficult enough to model such things when you have a handy extant species for comparison but for extinct populations the probability of arriving at a decent answer would seem to be vanishingly small.