Randomness in Evolution: Valid and Invalid Usage

[mijopaalmc]

Um, no mijo.

Um, yes.

Even if every phenotype has a equal probability of reproduction, the generalization of the central limit theorem still apply.

 

[Meadmaker]

I'd be satisfied if he'd simply admit that "mutations are random, but evolution isn't".

I prefer Sol's comparison to fire, in which he notes that saying "fire is random" is "not even wrong". The same is true for evolution. It is random, and it is not random, and any attempt to say that it must be called one or the other, absent any other context, is not even wrong.

The interesting thing to me is why you say you would be "satisified" if he were to "admit" that one side is right. That's what it's all about, really. It's about picking a side.

Of course, there is great concern that mijo, or others, are using the same words that THEY use, and that is an offense up with which we shall not put!

Except that, really, THEY don't use that word all that much. In the only example cited here, when they used it, they got it right.

 

[mijopaalmc]

I prefer Sol's comparison to fire, in which he notes that saying "fire is random" is "not even wrong". The same is true for evolution. It is random, and it is not random, and any attempt to say that it must be called one or the other, absent any other context, is not even wrong.

Actually, the problem seems to be that the majority of the phenomena that people who argue that evolution is non-random are not in and of themselves random, so the fact that we observe them in evolution by natural selection does not make it inherently non-random. By the way, my original argument, which seems to have been lost in the recesses of time, was that there is no evidence that evolution is non-random. Such a lack of evidence does not mean that evolution is therefore random, but my main point of confusion was why people were so adamant that evolution was non-random when the evidence they cited said nothing of the sort.

 

[sol invictus]

You and the bulk of those who argue that evolution is non-random make similar fundamental mistakes about probability theory (viz., articulett's argument that algebra is random by the definition I gave).

Why don't you give an example of a mistake I made, mijo?

I'm still waiting for a response.....

What's the matter - is there a problem?

Was that another lie, by any chance?

 

[Reality Check]

Actually, the problem seems to be that the majority of the phenomena that people who argue that evolution is non-random are not in and of themselves random, so the fact that we observe them in evolution by natural selection does not make it inherently non-random. By the way, my original argument, which seems to have been lost in the recesses of time, was that there is no evidence that evolution is non-random. Such a lack of evidence does not mean that evolution is therefore random, but my main point of confusion was why people were so adamant that evolution was non-random when the evidence they cited said nothing of the sort.

IMHO Convergent evolution is evidence that evolution is non-random.

 

[mijopaalmc]

I'm still waiting for a response.....

What's the matter - is there a problem?

Was that another lie, by any chance?

You are still focusing on the lack of certainty in the measurement of the initial conditions of system. Deterministic and stochastic process are mathematically distinct so they should at least in theory yield physically distinct results.

 

[mijopaalmc]

IMHO Convergent evolution is evidence that evolution is non-random.

What makes so sure that it can't happen if certain traits are favored over others?

 

[Reality Check]

What makes so sure that it can't happen if certain traits are favored over others?

What do you mean? What is doing the "favouring"?

 

[mijopaalmc]

What do you mean?

For instance, thick winter coat or layer of blubber is favorable to warm blooded animals who live in cold climates with long winters, because they don't have to consume as much energy keeping themselves warm. Therefore, we can reasonably expect to see animals evolve these and other energy conserving strategies (e.g., hibernation), if their environment get colder and has progressively longer winters.

What is doing the "favouring"?

Nothing per se is favoring these traits; it just that individual who possess them are more likely to survive and reproduce in the spring.

 

[Walter Wayne]

Um, yes.

Even if every phenotype has a equal probability of reproduction, the generalization of the central limit theorem still apply.

Central limit theorum applies to independent variables, each with the same distribution. The probability of a specific mutation happening is dependent on the current genomes. For example, mutation at a site from A->C might be very unlikely, but from B->C might be more likely.

If that is the case, then the likelihood of C arising is dependent on the gene frequencies in the previous generation. The central limit theorem will not apply to long term evolution in your hypothecal, unbias selection.

Walt

P.S. It's late so I can't really get to other posts. Another time.

 

[Reality Check]

For instance, thick winter coat or layer of blubber is favorable to warm blooded animals who live in cold climates with long winters, because they don't have to consume as much energy keeping themselves warm. Therefore, we can reasonably expect to see animals evolve these and other energy conserving strategies (e.g., hibernation), if their environment get colder and has progressively longer winters.

Nothing per se is favoring these traits; it just that individual who possess them are more likely to survive and reproduce in the spring.

So you agree that this is evidence that evolution is non-random?
Since similiar environments cause evolution of similiar traits then evolution is not random.

 

[mijopaalmc]

So you agree that this is evidence that evolution is non-random?
Since similiar environments cause evolution of similiar traits then evolution is not random.

No, just because an environment favors the development of a trait doesn't mean evolution is non-random. All the descriptions I have read of natural selection say that better-adapted individuals have a high probability of surviving than others. That is a description of a stochastic process.

 

[mijopaalmc]

Central limit theorum applies to independent variables, each with the same distribution. The probability of a specific mutation happening is dependent on the current genomes. For example, mutation at a site from A->C might be very unlikely, but from B->C might be more likely.

If that is the case, then the likelihood of C arising is dependent on the gene frequencies in the previous generation. The central limit theorem will not apply to long term evolution in your hypothecal, unbias selection.

Walt

P.S. It's late so I can't really get to other posts. Another time.

There are several central limit theorems that extended the convergence to normality variables that are dependent on each other.

 

[Reality Check]

No, just because an environment favors the development of a trait doesn't mean evolution is non-random. All the descriptions I have read of natural selection say that better-adapted individuals have a high probability of surviving than others. That is a description of a stochastic process.

Then this is a matter of semantics. When you say "non-random" you mean deterministic. When I say "non-random" I mean not random.

 

[mijopaalmc]

Then this is a matter of semantics. When you say "non-random" you mean deterministic. When I say "non-random" I mean not random.

That is a distinction without a difference. Something that is non-random is deterministic and vice versa.

 

[zosima]

I've got a few thoughts.

1. Regression of the mean works in opposition to evolution. It was one of the central reasons that lamarckian genetics was rejected. If everything could mix evenly any variation would regress back to the mean. Ie not net change over time. But in the discrete system of mendelian genetics that change cannot always regress. I may be explaining this point poorly so I encourage you to read about the history of the debate between the two.

2. I think macroscopic traits and complexity are often different things.
a. Evidence from the genome shows that we collect a lot of information in the genome, and the amount of information increases with time(just more junk in the genome). Evolution removes this sort of stuff, so it certainly decreases the complexity of the species(or gene pool) over time.(with a very strict definition of complexity)
b. Macroscopic traits like intelligence say very little about complexity. The way that we ended up being intelligent is probably not the simplest way to create a creature of equivalent complexity nor the most complex way. In other words evolution has a lot of appendices, this is particularly true in cognition, where we have newer layers of brain(in the evolutionary sense) layered on top of older ones and it seems a lot of those underlying layers do very little processing in humans. So a lot of extra complexity but very little extra intelligence.

3. Randomness is absolutely not necessary for evolution, but what you do need is variation. To have evolution in the way we generally think of it(ie not directed by things other than best fit to environment) that variation should be a good approximation of uniform variation. Luckily one of the constraints humans place on random variables(both intuitively and explicitly) is that those variables have a uniform distribution.

4. I personally think that the world is deterministic all the way down. As I understand quantum mechanics, the issue with hidden variables isn't that they are necessarily random, they can be non-random if they are non-local. I'm fine with non-locality, everything is connected, cool.

5. Given that, a fundamental question here is the definition of random. I think things are random if they appear random even if they have underlying deterministic behavior. If you disagree okay, but I think its a matter of opinion, or convention, if you will. I think once we finally define our terms this is a non-issue.

6. Finally there is a more interesting question beyond this. Does evolution necessarily converge on different forms? I can't say for any particular case, but what I can say for sure is that evolution is essentially a hill climbing algorithm(also often called simulated annealing). Which means it can only change from one form to another by small steps(or local variation). This precludes certain possibilities. So despite the fact that our brain is designed (I say designed only in the sense that I am anthropomorphizing an inanimate and non-directed process) by piling on layers, that now represent inefficiencies, it seems very likely that we couldn't have done it any other way. Any other intelligent evolved creatures in our universe would necessarily have their histories "built in" to their architecture, even if they end up doing it in a different way. It will never converge on the best way, because those intermediate steps are inseparable from the process( or algorithm). If you look at developmental biology all creatures on earth share this developmental commonality(and incidentally a common ancestor)

tl;dr

I'd just like to conclude with an interesting tangentially related thought. In A New Kind of Science, Wolfram claims there are actually 3 kinds of randomness rather than the two mentioned above.(This is actually the central claim of the book)

Specifically there are systems that don't have complexity built into their initial conditions(ie not chaotic systems), they don't have any "true" randomness, but instead they are systems that generate randomness.

They have simple homogeneous initial conditions and simple rules, but because they pass some incredibly low computational threshold they generate randomness. It's quite amazing.
http://mathworld.wolfram.com/Rule30.html
A 6 rule cellular automaton on a simple 1-d binary grid can generate a pattern that never repeats starting from a single black square. It is also proven that this system is a universal Turing machine. wow.

 

[Reality Check]

That is a distinction without a difference. Something that is non-random is deterministic and vice versa.

That is what I said: you say non-random instead of deterministic. I say non-random instead of not random.

 

[mijopaalmc]

That is what I said: you say non-random instead of deterministic. I say non-random instead of not random.

So are "non-random" and "deterministic" interchangeable in all cases?

 

[Reality Check]

So are "non-random" and "deterministic" interchangeable in all cases?

Your interpretation seems to be yes.
My interpretation is that a non-random process = everything that is not a random process. A deterministic process is a non-random process. A stochastic process is not a random process and so is a non-random process. Thus non-random" and "deterministic" are not interchangeable in all cases.

 

[mijopaalmc]

Your interpretation seems to be yes.
My interpretation is that a non-random process = everything that is not a random process. A deterministic process is a non-random process. A stochastic process is not a random process and so is a non-random process. Thus non-random" and "deterministic" are not interchangeable in all cases.

Unfortunately, a stochastic process is by definition a random process. If you are interested in learning the basics of what a stochastic process is, you should go to Wolfram MathWorld or PlanetMath an read the entries on stochastic process and follow the links to the terms you don't understand. Basically, any process that develops in time and is based on probabilities is a stochastic process.

The problem in discussing evolution by natural selection as a stochastic process is that there s very little evidence that demonstrates that individuals with identical phenotypes must all reproduce, making it a deterministic process. Most often individuals in evolution by natural selection are described as having a probability of reproducing, which is a description of a stochastic process.