Randomness in Evolution: Valid and Invalid Usage

[Paul C. Anagnostopoulos]

So you're saying that there is NOTHING non-random about selection ? Selection is random ?

If we insist on using the word random for every process except those that are 100% deterministic, then yes.

Do you find that confusing?

~~ Paul

 

[Paul C. Anagnostopoulos]

Evolution by natural selection is an orderly process that arises out of the random events of mutation and natural selection. There is nothing non-random about these events, but many repetitions lead tot he emergence of a pattern.

Would you be adverse to adding:

The pattern is statistically dependent on the inherited traits and the current environment.

If so, I think we can agree. Well, except that "nothing non-random" is misleading.

~~ Paul

 

[Belz...]

If we insist on using the word random for every process except those that are 100% deterministic, then yes.

Do you find that confusing?

No, just useless.

Random and deterministic need to mean something different, but they don't by Mijo's def.

 

[Wowbagger]

I'd like to thank all participants in making this my most popular thread, ever!

 

[mijopaalmc]

I'd like to thank all participants in making this my most popular thread, ever!

Well you certainly seem to know how to pick topics that spark lively debate, even if no-one concedes in the end.

 

[zosima]

Well you certainly seem to know how to pick topics that spark lively debate, even if no-one concedes in the end.

Think about it Mijo, you could concede, just ironically, and this debate would come to an end. You could bring a sense of resolution to literally 10s of lives!

lol

 

[articulett]

But then, that would ruin my prediction that Mijo must have the last word on this topic so that he can win the conversation in his head.

 

[Paul C. Anagnostopoulos]

So Mijo, would you be adverse to adding:

The pattern is statistically dependent on the inherited traits and the current environment.


~~ Paul

 

[mijopaalmc]

So Mijo, would you be adverse to adding:

The pattern is statistically dependent on the inherited traits and the current environment.

I believe I have said something like that before, although it is a very vague recollection. However, statistical dependence (or conditioning) does not imply nonrandomness.

 

[zosima]

I believe I have said something like that before, although it is a very vague recollection. However, statistical dependence (or conditioning) does not imply nonrandomness.

I think we can resolve this right now. I think most of us just think of random in a slightly different way. For example random in the context of computer science is most often defined as, uniformly distributed, uncorrelated, and unpredictable*. That ends up corresponding to an unbiased coin flip. It seems like your definition of random doesn't include the idea of uniformity. In fact, it allows probability distributions that are very skewed to be accepted as random under the definition. This would correspond more to a biased coin. The problem I think people have with the latter definition is that large skews undermine the idea of unpredictability. For example if a coin is heads 9/10 of the time and tails 1/10 of the time, I can predict fairly accurately that the coin will come up heads.

Could we agree that if we were to use the first definition then the path of evolution wouldn't be random? Even though you don't agree with that definition? But under the second definition the path of evolution would be random? Even though most of us don't agree with that definition?

In either case, no one is saying that the path of evolution is 100% deterministic. However,there seems to be a sub-argument about how skewed the probability distribution in evolution is. Some people, including myself, would assert that evolution is very close to deterministic, maybe skewed so that one outcome will be favored over another 99% or more of the time while all other outcomes will fall into that less than 1%, but thats not actually *that* central to the question of whether we call it random or not. I think what is central is whether you accept the first definition or the second.

*Whether unpredictable follows from the first two premises, or is necessary in its own right, is a matter of some dispute, and seems to follow from how you define correlation.

 

[mijopaalmc]

zosima-

I have already agreed several time that evolution by natural selection is non-random in so far as it is biased, correlated and predictable. The problem is that statistical hypothesis testing (even in maximum parsimony methods, especially when the number of taxa is above 8) used in evolutionary biology has a much wider usage of random that the aforementioned one, and it is therefore inconsistent to state that evolution by natural selection is non-random while making the assumptions of randomness necessary to perform any number of statistical hypothesis tests.

 

[zosima]

Question:

Could we agree that if we were to use the first definition then the path of evolution wouldn't be random? Even though you don't agree with that definition? But under the second definition the path of evolution would be random? Even though most of us don't agree with that definition?

Answer:

zosima-

I have already agreed several time that evolution by natural selection is non-random in so far as it is biased, correlated and predictable. The problem is that statistical hypothesis testing (even in maximum parsimony methods, especially when the number of taxa is above 8) used in evolutionary biology has a much wider usage of random that the aforementioned one, and it is therefore inconsistent to state that evolution by natural selection is non-random while making the assumptions of randomness necessary to perform any number of statistical hypothesis tests.

I'm confused, is that a yes? Or is that a no? And could you elaborate on your point wrt, statistical hypothesis testing? 'Cause if you mentioned it before you haven't really emphasized it.

 

[mijopaalmc]

Again, I agree that using the definition of "random" as "equiprobable" makes evolution by natural selection non-random.

I disagree I agree that using the definition of "random" as "equiprobable" is consistent with all the statistical assumptions that evolutionary biologists make to demonstrate evolution.

 

[zosima]

Again, I agree that using the definition of "random" as "equiprobable" makes evolution by natural selection non-random.

I disagree I agree that using the definition of "random" as "equiprobable" is consistent with all the statistical assumptions that evolutionary biologists make to demonstrate evolution.

Cool beans,That seems like a good compromise to me. I'm not sure what you mean about the statistical assumptions that evolutionary biologists make. I don't really understand the claim you're making and it seems like a slightly different tack than you've taken previously. So can you elaborate?

 

[Paul C. Anagnostopoulos]

I believe I have said something like that before, although it is a very vague recollection. However, statistical dependence (or conditioning) does not imply nonrandomness.

Agreed. However, the more skewed the probabilities, the more I think we're obliged to explain "evolution is random" in greater detail, so people understand in what way it is random. A die with a weight on the 4 side is not a "random die" in the sense that most people understand it.

Calling every naturalistic process "a stochastic process" does not reveal the true complexity of nature, even though it is true.

~~ Paul

 

[mijopaalmc]

Agreed. However, the more skewed the probabilities, the more I think we're obliged to explain "evolution is random" in greater detail, so people understand in what way it is random. A die with a weight on the 4 side is not a "random die" in the sense that most people understand it.

First off, bias is not the same thing as statistical dependence. Each roll of a weighted dies is independent of the rolls that come before and after it. The sum of a series of rolls of the same weighted die is dependent of the rolls that preceded it.

Second, I don't think the skew of the probabilities necessitate any longer an explanation than that of unskewed probabilities. For instance most people understand that, even though the probability of heads tossing a fair coin does not result in getting an exact 1:1 heads to tails ratio but that it is far more likely that the ratio will be between 1:.8 and 1:1.2 than it is to be between 1:.6 and 1:.8 or 1:1.2 and 1:1.4. Furthermore, they also understand that, as the number of tosses increases, it becomes more and more likely that the ratio will be between 1:.8 and 1:1.2 and less and less likely that it will be between 1:.6 and 1:.8 or 1:1.2 and 1:1.4. Thus the equidistribution of event probabilities does not guarantee the equidistribution of probabilities for the entire process.

Calling every naturalistic process "a stochastic process" does not reveal the true complexity of nature, even though it is true.

How does making a statement that, even if you can't be certain of every detail of a system, you can still describe it long-term orderly behavior not "reveal the true complexity of nature"?

 

[zosima]

The sum of a series of rolls of the same weighted die is dependent of the rolls that preceded it.

I'm confused as to what you're saying here. Are you saying we can't know the sum of a series of die rolls without knowing the outcome of each specific roll?
This seems trivially true. For sure we can determine the probability distribution of the sum of a series of rolls(weighted or unweighted).

How does making a statement that, even if you can't be certain of every detail of a system, you can still describe it long-term orderly behavior not "reveal the true complexity of nature"?

As with the previous statement, the idea that we can't be certain of every detail of any natural system is trivially true. I think people feel it is a poor way to phrase things because it is misleading. Particularly because on the level of description that is typically used in evolution many of those details we can't be sure of smear out. Do you understand why people feel this is a misleading statement?
For example the sum of a large number of weighted or unweighted die rolls is going to tend toward the expected value of those rolls times the number of rolls.
If we only care about the accuracy within say +-10% then we should be able to say what that sum is going to be with a very high confidence. This is often the case with the level of description in evolution. We can say with high confidence what the behavior of the variables we care about is going to be and the details we don't care about are inconsequential.

Also, you never clarified what you meant by "consistent with all the statistical assumptions that evolutionary biologists make to demonstrate evolution."
I still don't understand, what was the reasoning behind this statement? Did you miss my question or are you ignoring it? I meant it in earnest. It seems like you claimed that evolutionary biologists were making some major mistake but then failed to provide any reasoning for the assertion.

 

[mijopaalmc]

I'm confused as to what you're saying here. Are you saying we can't know the sum of a series of die rolls without knowing the outcome of each specific roll?
This seems trivially true. For sure we can determine the probability distribution of the sum of a series of rolls(weighted or unweighted).

For being such a trivial (or simple) truth, it sure has made people a lot of money in, for example, craps.

As with the previous statement, the idea that we can't be certain of every detail of any natural system is trivially true. I think people feel it is a poor way to phrase things because it is misleading. Particularly because on the level of description that is typically used in evolution many of those details we can't be sure of smear out. Do you understand why people feel this is a misleading statement?
For example the sum of a large number of weighted or unweighted die rolls is going to tend toward the expected value of those rolls times the number of rolls.
If we only care about the accuracy within say +-10% then we should be able to say what that sum is going to be with a very high confidence. This is often the case with the level of description in evolution. We can say with high confidence what the behavior of the variables we care about is going to be and the details we don't care about are inconsequential.

Also, you never clarified what you meant by "consistent with all the statistical assumptions that evolutionary biologists make to demonstrate evolution."
I still don't understand, what was the reasoning behind this statement? Did you miss my question or are you ignoring it? I meant it in earnest. It seems like you claimed that evolutionary biologists were making some major mistake but then failed to provide any reasoning for the assertion.

You have just summarized the essence of a statistical statement; you can say with a certain level of confidence, but not complete confidence, that a statement is true. This is the basis for all statistical tests (including those used by evolutionary biologists to confirm their hypotheses); a test statistic is generated for sample data and the compared to a distribution of test statistics to determine whether the differences within or among samples are dues to the inherent variability of the sample or due to fact that the samples really are different.

 

[zosima]

....

You ignored all my questions.
In lieu of answers, I assume:
1. You don't understand what anyone here is talking about.
2. That your claim about the use of statistics in evolutionary biology is made up and without merit.
3. That you are ignoring my questions about your claim because it is easier than admitting you made it up.

If you choose to correct my assumptions, I would appreciate it. Otherwise, I can sleep happy knowing this issue is settled.

 

[mijopaalmc]

You ignored all my questions.
In lieu of answers, I assume:
1. You don't understand what anyone here is talking about.
2. That your claim about the use of statistics in evolutionary biology is made up and without merit.
3. That you are ignoring my questions about your claim because it is easier than admitting you made it up.

If you choose to correct my assumptions, I would appreciate it. Otherwise, I can sleep happy knowing this issue is settled.

And you have ignored that I actually did answer your questions. Your assumption are therefore an reflection of your own unwillingness to attempt to understand.