Randomness in Evolution: Valid and Invalid Usage

[Wowbagger]

I, on the other hand, think that the way evolutionary biologists describe the process of evolution by natural selection as it occurs in the physical world is inherently stochastic (and not necessarily just because of out lack of knowledge of the details of the process) and that the existing data that we have corroborates this interpretation.

How do you know? Stochastic implies that the system is ultimately deterministic, anyway. We just can't determine everything, because our knowledge is imperfect.

If you think the nature of Evolution is fundamentally stochastic, you are going to have to find a source or mechanism of indeterminism, in the system (other than quantum uncertainty, which is largely wiped out in the scales Evolution works in).

Until such a new source of "random" fluctuations are discovered, it seems best to think of Evolution as ultimately deterministic, so that we may continue making discoveries and formulating predictions with it.

You seem to have a very strong mathematical background, mijo. Perhaps your training is making you forget that science is based on evidential support, not what someone thinks.

There are non-chaotic physical systems where this isn't the case. The key word is "significantly".

Ah, good to emphasize the key words.

I would argue that this means that the selection pressures, and thus the "direction" of evolution is subject to change due to random events.

If the direction is subjected to random change, then surely this is "random".

It might be a matter of resolution. In the large scales Evolutionary models work in, such random events usually have a miniscule impact on the system, if any at all. The patterns of convergence we see is testimony to that.

Over long timescales, with unstable environments, I would argue that the random element is important, as it will alter the direction of the selection pressures. This is important if one is discussing the evoultion of humanity, as the KT impact, the ice-ages, the cambrian extinction, Toba, etc could have altered the course of evolution so that intelligent hominids wouldn't evolve, maybe so that nothing occupied the ecological niche of complex-tool and fire using social animals.

But your examples (the KT impact, ice ages, etc.) were large enough events that any small changes in random fluctuation would not have stopped them from happening. In that sense, they are NOT random.
(Though, in another sense, you could use random to mean "unanticipated" by the life forms on the planet. But, such usage is probably not useful in describing the Evolutionary process.)

 

[jimbob]

Wowbagger,

I think we are emphasising slightly different things here. I also suspect that I have a lower "threshold of significance" than you, i.e. you are looking very broad-brush, whilst I am looking at the level of particular niches:

There are many pressures that are almost univeral and which act upon every member of say a kingdom, maybe even higher.

All animals will need to eat; they are unlikely to be able to afford to "waste" energy, so there will always be a pressure in favour of "frugality". Sometimes there are also opposing pressures, for example towards flight, however the ability to fly has a cost in energy, so should the need for flight vanish, the "frugality" pressure will act to reduce the ability to fly.

Is this the level of "significance" that you are talking about? (or indeed, the level of resolution that you are talking about).

At that resolution I would agree with you, however I am thinking about a *slightly* finer resolution.

Devil's Advocate: "Suppose our ancestors went through a stage of scavenging in the savannah, and in a slightly different universe, there was an animal that was a lot better at this than them, but that there was slightly less competition back in the trees. The surviving hominids could have ended up occupying similar niches to the chimps or bonobos, and not as ancestors to fire-using social animals."

An entire niche could remain unocupied/unavailable due to a slightly different set of competing organisms due to a different order of mutations altering the fitness landscape for other organisms.

This above example isn't terribly clear, I'm afraid, but I hope you can see what I am getting at.

 

[jimbob]

Maybe a clearer example would be to say that the rhino and triceretops seem to occupy/have occupied similar niches (ones that are highly likely to be filled). But I would argue that there are significant differences, and many of them are due to the randomness of the process.

 

[Walter Wayne]

Wowbagger,

Stochastic does not mean ultimately determistic.

Of course mijo is confusing the model with reality. The use of a stochastic model imples that the process is random or that it has hidden/unknown variables. So the use of the stochastic model doesn't tell us much about the randomness or lack of thesystem.

Walt

 

[mijopaalmc]

Since evolution is a process, I don't see how they could mean different things. But I'm willing to learn.

~~ Paul

I was thinking along the lines of the idea that bird dog doesn't mean a bird that is also a dog, or vice versa. In other words, maybe the phrase "random process" means something different than the sum of the words "random" and "process". I asked this question partially because "random process" has a exact and rigorous definition within probability theory and it therefore avoid the definitional problems that people insist on have with the single word "random".

 

[mijopaalmc]

I said it before and I'll say it again:

Variation is non-deterministic with respect to form.
Selection is deterministic with respect to form.

It does not matter ONE BIT whether or not that variation is caused by quantum mechanics, a die roll, a list of numbers, or even, horrors, a deterministic process, etc.. the important part is the deterministic relationships.

If variation is based on form then mutation is not the mechanism of change in form - previous forms are.

If selection is not based on form then what variations arise are inconsequential to what forms persist.

These are the basic and fundamental conditions that must be satisfied for any system that would be considered "evolutionary" - be it biological or a abstract model.

NOTA BENE deterministic with respect to is not the same as wholly determined by - yes mijo, I am thinking of you and your twins. Their form does not wholly dictate their selection but to argue that their selection is not deterministic with respect to it would be WRONG.

You are choosing to redefine the words "random" and "deterministic" to preclude evolution from being random, cyborg. In fact, you are ignoring that even if there were no bias in selection and no novel mutations introduced into the population that evolution would still occur purely by regression toward the mean. That is why you are wrong.

 

[Wowbagger]

I think we are emphasising slightly different things here. I also suspect that I have a lower "threshold of significance" than you, i.e. you are looking very broad-brush, whilst I am looking at the level of particular niches:

The threshold is variable. And, yes, perhaps we are both right, but in different parts of the range.

Devil's Advocate: "Suppose our ancestors went through a stage of scavenging in the savannah, and in a slightly different universe, there was an animal that was a lot better at this than them, but that there was slightly less competition back in the trees. The surviving hominids could have ended up occupying similar niches to the chimps or bonobos, and not as ancestors to fire-using social animals."

No need for DA tags, in that one. I would agree with it.

But, think about how much different the world would have needed to be, in order for that alternative scenario to take place. We are no longer talking about differences from mere randomness, there.

A bucket of water with one less drop in it, is still a bucket of water, in the end.

The flapping of a butterfly's wings may contribute to a hurricane, but would never, by themselves, generate one.

An entire niche could remain unocupied/unavailable due to a slightly different set of competing organisms due to a different order of mutations altering the fitness landscape for other organisms.

It is possible, I suppose. But, I wonder: How often are slight differences really responsible for such things?

Can anyone offer examples where some slight difference clearly did make a huge impact on the course of evolution for some life form, that would not otherwise have occurred?

Maybe a clearer example would be to say that the rhino and triceretops seem to occupy/have occupied similar niches (ones that are highly likely to be filled). But I would argue that there are significant differences, and many of them are due to the randomness of the process.

IIRC, Kangaroos and antelopes are another example of similar niches being filled by different-looking animals.

But, no matter what the details, we could still have predicted the niche would be filled.

Stochastic does not mean ultimately determistic.

Well, perhaps, strictly speaking, yes. However...

The use of a stochastic model imples that the process is random or that it has hidden/unknown variables. So the use of the stochastic model doesn't tell us much about the randomness or lack of thesystem.

...for scientific purposes, it seems Occam's Razor would have determinism as the simplest explanation, unless evidence otherwise (with Heisenberg-like precision) is found.

If mijo thinks Evolution is fundamentally random, he has to deliver such evidence.

 

[mijopaalmc]

Congratulations, you managed to find merely the most commonly used definition of "random" (and probably the only definition of "random" used on those sites) and are now trying to pass it off as only definition in use in the English language.

So mijo, you really don't remember that long and excruciatingly annoying discussion we had in the "Evolution Not Random" thread about how your definition is useless, since according to it every single process in the world is random?

Strange, because here you are in a different thread more recently parroting almost verbatim what I was telling you there all along:

Except, to the best our knowledge, quantum mechanical events are random no matter how you try to constrain the variables. While it is mathematically true that a function of a random variable is itself a random variable*, it is not very useful in describing real world physical processes, because every actual physical process is based on quantum mechanics. However, it still important to make a distinction between non-chaotic deterministic systems (i.e., roughly systems in which small variations in initial conditions lead to small variations in final conditions) and chaotic deterministic systems (i.e., roughly systems in which small variations in initial conditions lead to large variations in final conditions). Evolution is most likely the latter because it takes place in a in an environment that is dependent on other chaotic systems (e.g., climate and weather). This means that exact predictions cannot be made about specific elements of adaptation, whereas predictions about the general course of adaptation can be made and are best handled in a probabilistic/statistical framework.

*meaning that random mutation based on probability leads to random evolution by natural selection regardless of whether natural selection itself is random or deterministic

I think you are misunderstanding what I wrote here. I was mainly responding to the fact that I had made the claim earlier in the thread that a function of a random variable (e.g., genomes after mutations) is a random variable, and evolution is therefore a random process because mutation is random. Consequently, I was saying that, even though this is incontrovertibly true from a mathematical standpoint, it is rather useless in describing physical systems if all our measurements are in fact random variables described by probability distributions.

The second part is a bit harder to understand because I seem to be describing a chaotic process in the way in which I would normally describe a stochastic process. However, my point was and still is, if our measurement are inherently imprecise and evolution by natural selection is indeed a chaotic system, it does us little good to think of it as a deterministic system (even though it is by definition deterministic), because (and this is part of the mathematical definition of a chaotic system) no matter how exact our measurements and how complete our knowledge of the systems are there will be some areas of the state space where a small differences in initial conditions are going to lead to large differences in final conditions. Therefore, it is much more practical and precise, due to the chaotic nature of the system, to derive a probability distribution for the possible values the measurements could assume.

 

[mijopaalmc]

Wowbagger-

How is assuming randomness less parsimonious that assuming determinism?

 

[Wowbagger]

How is assuming randomness less parsimonious that assuming determinism?

Randomness lies outside of experimental science. A process that is truly random cannot be recreated in the laboratory, (at least not with any consistency).


ETA: Another, possibly more fitting, answer: The mechanism of randomness would have to be explained, and without evidence of its presence, should be disgarded. Natural determinism in a process needs no extraneous explanation.

 

[mijopaalmc]

Randomness lies outside of experimental science. A process that is truly random cannot be recreated in the laboratory, (at least not with any consistency).

You're missing the point yet again. Randomness can be tested in the lab (and in fact been tested to an absurdly high degree in the case of quantum mechanics).

ETA: Another, possibly more fitting, answer: The mechanism of randomness would have to be explained, and without evidence of its presence, should be disgarded. Natural determinism in a process needs no extraneous explanation.

Evidence of randomness's presence is identical initial conditions yielding different final conditions. This is how randomness is defined mathematically and scientifically. you are failing quite spectacularly at understanding this.

 

[Wowbagger]

Randomness can be tested in the lab (and in fact been tested to an absurdly high degree in the case of quantum mechanics).

I agree with that. But, you have yet to show us such a source of randomness in the nature of Evolution, that can be tested in the lab.

Quantum mechanics has certainly been thoroughly tested, but its impact is largely erased at the large scales the process of Evolution works in.

Without evidence of a new source of randomness, it is better for the sake of scientific investigation, to assume one does not exist.

Evidence of randomness's presence is identical initial conditions yielding different final conditions.

I would love to see some examples of this, in Evolution.

This is how randomness is defined mathematically and scientifically. you are failing quite spectacularly at understanding this.

This topic is about how the word Random applies (or does not apply) to Evolution. How are those definitions relevant to Evolution, then?

I suspect that your mathematical training might be keeping you blind to the importance of evidential support.

 

[mijopaalmc]

Wowbagger-

Are you trying to tell me that natural selection neatly divides a population into two collections of phenotypes: one where no individual who possesses one of the constituent phenotypes has reproductively viable offspring and the other where every individual who possesses one of the constituent phenotypes has reproductively viable offspring?

That is what I hear when someone claims that evolution by natural selection is a deterministic process, and that is the kind of evidence that someone who claim that evolution by natural selection is a deterministic process.

By the way, I do understand that evolution by natural selection is non-random (i.e., deterministic) by the way that evolutionary biologist choose to define it, but I think evolutionary biologists' choosing to describe the actual process in this way is extremely inconsistent with the other ways in which evolutionary biologists use the concept of "random" within their own field.

 

[Taffer]

Wowbagger-

Are you trying to tell me that natural selection neatly divides a population into two collections of phenotypes: one where no individual who possesses one of the constituent phenotypes has reproductively viable offspring and the other where every individual who possesses one of the constituent phenotypes has reproductively viable offspring?

That is what I hear when someone claims that evolution by natural selection is a deterministic process, and that is the kind of evidence that someone who claim that evolution by natural selection is a deterministic process.

By the way, I do understand that evolution by natural selection is non-random (i.e., deterministic) by the way that evolutionary biologist choose to define it, but I think evolutionary biologists' choosing to describe the actual process in this way is extremely inconsistent with the other ways in which evolutionary biologists use the concept of "random" within their own field.

You keep saying what evolutionary biologists do and do not. Can you please elaborate on how these biologists use inconsistant definitions of random?

 

[mijopaalmc]

You keep saying what evolutionary biologists do and do not. Can you please elaborate on how these biologists use inconsistant definitions of random?

Again evolutionary biologists use an implicit understanding of randomness that is purely based in probability whenever the perform statistical hypothesis testing. This is especially true if they use parametric statistics, which require them to assume that population is described by a specific distribution (determined by the test they are performing) and test the sample data against that assumed distribution. It should not be hard to see how such a perception of randomness is inconsistent with any of the multitude of definitions that people have provide over the last 10 months.

 

[Taffer]

Again evolutionary biologists use an implicit understanding of randomness that is purely based in probability whenever the perform statistical hypothesis testing. This is especially true if they use parametric statistics, which require them to assume that population is described by a specific distribution (determined by the test they are performing) and test the sample data against that assumed distribution. It should not be hard to see how such a perception of randomness is inconsistent with any of the multitude of definitions that people have provide over the last 10 months.

Firstly, it's odd because a specific definition of "random" never came up when I was performing multitudes of population genetics and phylogenetics analyses. Secondly, what matters in these cases is what randomness causes, not what causes randomness. In other words, it doesn't matter if "random" means "acausal" or "deterministic but chaotic" or "goddidit". The tests work either way as long as the random null hypothesis is the same.

 

[Wowbagger]

Are you trying to tell me that natural selection neatly divides a population into two collections of phenotypes: one where no individual who possesses one of the constituent phenotypes has reproductively viable offspring and the other where every individual who possesses one of the constituent phenotypes has reproductively viable offspring?

I am not sure I understand the question. But, I will do my best to address what it seems to be addressing:

First of all: the reproductive success of a phenotype depends on the fitness landscape. In some landscapes a phenotype will result in much better survival, in others it could instantly kill the life form. In most situations, it will probably lay anywhere in between. (Keep in mind that the life form's adapted survival strategies and evolutionary heritage are factored into this, as well.)

Second of all: I very much doubt natural selection neatly divides anything into two collections. Nature tends to abhor taxonomy. There will always tend to be gray areas between two extremes. (The appearance of neat categories is the fault of both human bias in perception, and the natural tendency for phase-shifts and clumping to occur in some of the data.)

That is what I hear when someone claims that evolution by natural selection is a deterministic process, and that is the kind of evidence that someone who claim that evolution by natural selection is a deterministic process.

You probably heard incorrectly.

It is possible for nature to be both deterministic, and fail to neatly divide things into discrete categories.

It is humans who need to taxonomize, not nature.

By the way, I do understand that evolution by natural selection is non-random (i.e., deterministic) by the way that evolutionary biologist choose to define it, but I think evolutionary biologists' choosing to describe the actual process in this way is extremely inconsistent with the other ways in which evolutionary biologists use the concept of "random" within their own field.

It might be true that the word "random" is not used consistently throughout the field. (That is why I have so many bullet points in the opening post.) Other words, such as "gene" and "species" are also inconsistently defined, in the field. However, it is important for the working usage to be consistent within the same study. As long as each individual study uses words consistently in themselves, this problem is manageable.

I would certainly like to see words used even more consistently, myself. But, there is little the two of us can do about it. Especially since science is in the business of altering its models. As the models improve, definitions get updated. Even if everyone now uses the words in exactly the same way, it will not stay that way forever.

 

[sol invictus]

I think you are misunderstanding what I wrote here. I was mainly responding to the fact that I had made the claim earlier in the thread that a function of a random variable (e.g., genomes after mutations) is a random variable, and evolution is therefore a random process because mutation is random. Consequently, I was saying that, even though this is incontrovertibly true from a mathematical standpoint, it is rather useless in describing physical systems if all our measurements are in fact random variables described by probability distributions.

The second part is a bit harder to understand because I seem to be describing a chaotic process in the way in which I would normally describe a stochastic process. However, my point was and still is, if our measurement are inherently imprecise and evolution by natural selection is indeed a chaotic system, it does us little good to think of it as a deterministic system (even though it is by definition deterministic), because (and this is part of the mathematical definition of a chaotic system) no matter how exact our measurements and how complete our knowledge of the systems are there will be some areas of the state space where a small differences in initial conditions are going to lead to large differences in final conditions. Therefore, it is much more practical and precise, due to the chaotic nature of the system, to derive a probability distribution for the possible values the measurements could assume.

It doesn't sound like I misunderstood you at all - what you just said in the quote above is what I spent quite a number of posts trying to explain to you in the previous thread. I'm glad I got through.

So it sounds like we agree - your definition of random is useless since it defines all physical processes as random, and hence the statment "evolution is random" is empty. More useful definitions exist (and in fact are the ones used by everyone but you), and according to most of those, evolution is not random.

 

[articulett]

'Just to weigh-in on this thread... I want to know if jim-bob, walter wayne, or mijo ever make sense or cede a point. I have a feeling that they'll be arguing that it makes sense to call evolution "random" as long as they are forum members... and I have my strong suspicions as to why. Jimbob has his own weird definition that he thinks is fabu (random mutation and probabilistic selection which is almost as empty as mijo's bizarre insistence at calling selection "random" because it can be described in terms of probabilities before the fact...) And I never could figure out Walter Wayne. I've ended up putting them all on ignore figuring that someone would quote them if they said something intelligible... but so far, I'm not seeing any intelligence or comprehension... just their same old insistence that it makes sense somehow to describe evolution in the same manner that Behe describes it no matter how misleading or empty such a definition is.

You have to admire their tenacity, however,-- in fact, it reminds me Behe.

Wowbagger and Sol Invictus... mark my words... you are having a discussion with people who have a vested interest in not understanding. No progress will be made. It's not you... it's not your explanations; it's them. These guys were saying the exact same nothingness over a year ago on Mijo's "evolution is not nonrandom" thread"... truly... nothing has changed. Nothing.

 

[Walter Wayne]

So it sounds like we agree - your definition of random is useless since it defines all physical processes as random, and hence the statment "evolution is random" is empty. More useful definitions exist (and in fact are the ones used by everyone but you), and according to most of those, evolution is not random.

I agree with that mijo definition is useless, but I'm am wondering why you think by most other definitions evolution is not random, or at least which of those definitions would you think would not apply?

Walt