What evidence is there for evolution being non-random?

[EugeneAxeman]

There is a flawed assumption that the traits have foreknowledge of the conditions which will favor their expression.

More likely is that these recessive traits occassionally assert themselves, creating mutations which cannot survive unless the specific favorable environmental conditions exist.

It's similar to roulette.

It's possible for 17-black to come up, but if nobody plays it, that fact is irrelevant.

 

[Dr Adequate]

You are entirely welcome to your belief in intelligent design.

Just realize that you can't prove the intelligence that exists behind it.

Er ... you may have completely misread Taffer's post.

Either that or you meant to reply to someone else.

 

[Meadmaker]

Look here for example. "Random number", "random matrix", "random walk", et cetera, are all defined. "Random" is not.

Did you check page 2 of your link?




(To be fair, it would be more precise to say that evolution is a random process, characterized by random variables, etc.)

 

[Dr Adequate]

Did you check page 2 of your link?

Yes, but I missed it.

OK, it's a synonym for "stochastic".

And stochastic means?

"Stochastic is synonymous with "random." The word is of Greek origin and means "pertaining to chance" (Parzen 1962, p. 7). It is used to indicate that a particular subject is seen from point of view of randomness. Stochastic is often used as counterpart of the word "deterministic," which means that random phenomena are not involved."

That is not precise, nor technical, nor without ambiguity.

 

[mijopaalmc]

Yes, but I missed it.

OK, it's a synonym for "stochastic".

And stochastic means?

"Stochastic is synonymous with "random." The word is of Greek origin and means "pertaining to chance" (Parzen 1962, p. 7). It is used to indicate that a particular subject is seen from point of view of randomness. Stochastic is often used as counterpart of the word "deterministic," which means that random phenomena are not involved."

That is not precise, nor technical, nor without ambiguity.

It's interesting how you left out the last sentence, which I quoted and emphasized at least once before:

Therefore, stochastic models are based on random trials, while deterministic models always produce the same output for a given starting condition.

I hope that clears things up.

 

[Meadmaker]

That is not precise, nor technical, nor without ambiguity.

Ok. Guilty of oversimplification. Evolution is a random process. A random process is one whose state variables are described by random numbers.

 

[mijopaalmc]

Ok. Guilty of oversimplification. Evolution is a random process. A random process is one whose state variables are described by random numbers.

But the state variables are not random numbers in the same way that a random number is described:

A random number is a number chosen as if by chance from some specified distribution such that selection of a large set of these numbers reproduces the underlying distribution. Almost always, such numbers are also required to be independent, so that there are no correlations between successive numbers.

 

[Taffer]

The T-Rexes couldn't predict it, that's for certain.

So?

This goes to the nature of randomness and our universe. Is anything truly random? If we knew the initial conditions, could we predict all future activity? Does God play with dice?

The current general belief is that God does play with dice. There are events that occur in our universe that would have occurred differently but for the quantum level randomness that is present. That asteroid was not fated to hit the Earth, and its impact is best understood as a random event. Given a little bit of different dice jiggling during the formation of iron molecules in whatever star produced the particles that became the asteroid, it could have missed.

At least, that's the way an awful lot of scientists see things.

Erm, no. I'm fairly certain that cosmologists and astronomers would disagree with you. Once the properties of the asteroid were know, it could be accurately predicted. Ditto, once the properties (read: variation) of a population is known, the direction selection will take it can be accurately predicted. This is why there is a field known as 'population and evolutionary genetics'.

 

[Meadmaker]

But the state variables are not random numbers in the same way that a random number is described:

Oh all right. The state variables are random variables. The specific outcome of any experimental trial is found by selecting a random number from a distribution that has the same underlying probability distribution function as the corresponding random variable.

That does it! Screw precision! I'm saying "It's random!"

Seriously, though, what interests me is why people think it's misleading to call evolution "random", with or without all the high fallutin' words like "probability distribution function" that I learned in math class.

 

[Taffer]

Did you check page 2 of your link?

(To be fair, it would be more precise to say that evolution is a random process, characterized by random variables, etc.)

To post this before we fire off endless rebuttles to each other before coming to the point: I am more comfortable with this. A random process need not include entirely random elements. I maintain that selection is non-random. This is because we can both model it, and because it is necessary for evolution to occur. Just random mutation is not enough to drive evolution. This is the key misunderstanding with most creationists.

As I said earlier, I would always describe evolution as: "Evolution is the result of non-random selection acting upon randomly generated variation within a population."

 

[Taffer]

Oh all right. The state variables are random variables. The specific outcome of any experimental trial is found by selecting a random number from a distribution that has the same underlying probability distribution function as the corresponding random variable.

That does it! Screw precision! I'm saying "It's random!"

Seriously, though, what interests me is why people think it's misleading to call evolution "random", with or without all the high fallutin' words like "probability distribution function" that I learned in math class.

Because evolution is not just random, and to call it so is misleading, and leads to such misunderstandings as those that generate creationists' arguments.

 

[Meadmaker]

I looked up the whole asteroid question. I can say a couple of things with certainty. (No random numbers here.)

1. Way back in non-linear dynamics class, I remember in the explanation of chaotic dynamics that the professor said that even if you knew the initial position of a chaotic system (his example was the break of a set of billiard balls) to the limit imposed by the Heisenberg Uncertainty principle, you wouldn't be able to predict where the billiard balls would end up. I infer that quantum variation is enough to result in macroscopically observable differences.

2. The prevailing theory of why the universe is nonuniform, with galaxies, galaxy clusters, etc, involves quantum fluctuations during the expansion phase immediately following the big bang. I infer that if quantum fluctuations can move a galaxy, they can move an asteroid.

So?



Erm, no. I'm fairly certain that cosmologists and astronomers would disagree with you. Once the properties of the asteroid were know, it could be accurately predicted.

But the asteroid didn't start out as an asteroid. It started out as hydrogen atoms, that fused to helium atoms, that eventually became iron and other atoms, that attached to each other as dust, that gradually accumulated into pebbles, that stuck together into a really big rock. I'm fairly certain that somewhere along the line, something happened that might have resulted in that asteroid being a few meters away from where it was, and after a few billion years in orbit, that few meters could grow to a few thousand kilometers, and that would be enough to make it miss, at least on that particular orbit.

Ditto, once the properties (read: variation) of a population is known, the direction selection will take it can be accurately predicted.

Not if it gets hit by an asteroid.

 

[Taffer]

Consider two (rather imprecise) descriptions of evolution:

In each generation of organisms, some have traits that make them less fit for their environment, and some have traits that make them more fit. The ones that are more fit pass more of their genes on to the next generation. In each generation, there is also a certain amount of variation introduced, in the form of mutations. Some of these mutations make the organism more fit for its environment, some less. Those that are more fit will pass them on. Over time, those that make the organism more fit become more numerous, while the less fit ones pass out of existence.

Some variations also make an organism able to survive in different environments. This can allow organisms to expand their range or occupy vacant ecological niches. In response to selection pressure, the organisms and their genes become optimized to fill their specific niches in the environment.

Vs.


In each generation, some organisms survive, and some do not. Those that happen to survive pass their genes to the next generation. Since some genes are more likely to survive than others, over time those genes will become more numerous. In each generation, there are also imperfect copies of genes introduced. Some of those may have a survival benefit for an organism, and the imperfect copies may actually eventually become more numerous through having greater survival rates.

Some of those variations on the original genes may make an organism more likely to survive in a slightly different environment. A change in the environment an organism occupies might result in a change in which genes are more likely to survive, thus triggering a change in the population of the gene frequencies. As organisms occupy different environments, the different survival rates can create a higher variation in the range of observed gene frequencies.


The two descriptions say basically the same thing, but I suspect people will be attracted to one or the other. The first one describes “fitness”, “optimization” “selection”. The second emphasizes randomness, through phrases like “happen to survive”, “more likely”, “survival rate”, “higher variation”. The first says that life “expands” to fill niches. The second suggests that environments change, forcing change onto the life within it. I think the second is less misleading than the first. The first seems to suggest a purpose, a goal, a method. The second suggests random variation that happens to work out sometimes. I think that’s a better description of evolution.

Uncertain does not, in my opinion, mean truely random. And FWIW, these 'random' variations you describe are modeled (to the best that they can) in population models, and given the name "genetic drift" (yes, yes, oversimplified. So sue me.). But selection itself, not whether some organisms live and some die, not whether asteroids hit the earth, not whether a lizard eats a bird's egg, selection alone is not random. It acts the same way on all things, and is completely deterministic.

 

[Meadmaker]

Just random mutation is not enough to drive evolution. This is the key misunderstanding with most creationists.

Perhaps, but do you have a reference for any specific creationist that made this key misunderstanding?

 

[Taffer]

I looked up the whole asteroid question. I can say a couple of things with certainty. (No random numbers here.)

1. Way back in non-linear dynamics class, I remember in the explanation of chaotic dynamics that the professor said that even if you knew the initial position of a chaotic system (his example was the break of a set of billiard balls) to the limit imposed by the Heisenberg Uncertainty principle, you wouldn't be able to predict where the billiard balls would end up. I infer that quantum variation is enough to result in macroscopically observable differences.

2. The prevailing theory of why the universe is nonuniform, with galaxies, galaxy clusters, etc, involves quantum fluctuations during the expansion phase immediately following the big bang. I infer that if quantum fluctuations can move a galaxy, they can move an asteroid.



But the asteroid didn't start out as an asteroid. It started out as hydrogen atoms, that fused to helium atoms, that eventually became iron and other atoms, that attached to each other as dust, that gradually accumulated into pebbles, that stuck together into a really big rock. I'm fairly certain that somewhere along the line, something happened that might have resulted in that asteroid being a few meters away from where it was, and after a few billion years in orbit, that few meters could grow to a few thousand kilometers, and that would be enough to make it miss, at least on that particular orbit.



Not if it gets hit by an asteroid.

Ok, granted. I do not confess to be a math student (squishy science for me! ). I was in error.

 

[Taffer]

Perhaps, but do you have a reference for any specific creationist that made this key misunderstanding?

You are kidding, right? "There is no way that randomly assembling parts can give rise to animals, much that a 747 cannot randomly assemble itself." A paraphrase, I know, but I'm sure no-one would argue that no creationist has ever used that argument. It stems from a key misunderstanding of evolution.

 

[Taffer]

Celestial mechanics is very complex, the three body solution is fairly moderate in complexity, the four body is much more difficult and the five body equations are intense.(At least that is what my math friends have told me.)

The level of accuracy is usualy more like "Within a magin of 300,000" miles in some cases and "within a margin of 60,000" miles for others. The level of accuracy is somehat vauge and fuzzy. You can not take an asteroid and predict to a great accuracy where it will be in fivehundred years. The level of precision is rather low.

Ok, I confess my error. It doesn't really change my argument, though. Selection, in and of itself, is not random. That that is the key to why evolution works.

And rainfall is totaly deterministic. as is food source production and growth, as well as parasites and disease. When you right your program that predicts the future let us know will you?

I never said we could right now. Please excuse my philosophy minor coming to the fore. It is possible to, given technology, knowledge, etc. I.e. I do not consider those things 'truely random'. Although in this, as above, I could be in error. I do not confess to be a physicist either.

Not even close, define accuracy as what, a vauge aprroximation that has no bearing on reality. So farmers can use these models to say where bugs will infest thier crops, or where to preserve the endagered speies to a few acres?

I am a determinist. Excluding 'truely random' quantum events, I consider 'macro events' to be understandable and modellable. We mightn't be able to do so right now, however.

Demonstarte a model with a level of precision, are you sure you have looked at the success of models? can you say where the next outbreal of longhorn beetles will be, can you narrow it down to specific counies, why or why not? You can save the timber industry a fortune.

Again, I never claimed it was possible to do all this right now. It is possible to do so, but not necessarily right now.

And when you demonstrate a model that works I might agree with you. So you can tell me how many tons of soy beans will be in the new crop on October 15 of 2007? That would be really helpful to the comodity market.

It doesn't matter. It is misleading to call evolution random. Why would we bother to model it, if it was? Why are we careful to teach genetics undergrads that evolution is non-random selection acting on random variation in a population? Why are we so defiant when creationists make strawman arguments which claim that evolution is impossible because "it is random"? It is because selection is not random.

What is the difference between a chaotic system and a random system, what meaning can be assigned to one that can not be assigned to the other. Especialy a chaotic system with twenty or more variables and millions of actors?

I would call a chaotic system one which is possible to model, whether or not we can right now, and a random system one which is not.

I think you over estimate the meaning of determined, the cause effect cycle may be known, or speculated upon, but that does not mean that it is deterministic.

Deterministic can simply be described as "all causes have an effect".

Tell me how many head of cattle will be slaughtered on July 15 of 2007?

If I knew all variables involved, and had models for the interaction of all these variables, then I believe I could.

 

[articulett]

What would it take, then?

I notice he didn't answer. I still contend that he's a creationist though he denies it. I predict that this thread will be like his fossil thread where people bend over backwards to explain and illustrate a somewhat simple concept with multiple links and examples which he'll ignore and dismiss and then tell everyone that they didn't answer his question, and so evolution is really random and he'llo just ask his biologist friend to explain it better.

It's simple if you aren't a creationist.

Mutations are "random" (more or less)--
Selection is not (more or less)--the specifics (more are less) are not important until you get the basics. There are trillions of experiments going on all the time. Only a very small percentage of life lives long to reproduce the another generation--some more successfully than others.

The dogs we have we breed from a very small group of wolves--bred for traits we humans like--over about 12000 years. We aided the survival and reproductive success of the dog creatures we valued the most...or at least the ones that irritated us the least were allowed to feast upon our historical leftovers...where their more timid or aggressive kin might not have been allowed the privilege.

Nature does the same kind of thing--but it takes a lot longer. The results are all the fabulously cobbled together creatures and genomes you see before you.

Mijo seems to purposefully obfuscate things to keep himself from understanding this simple notion as far as I can tell.

 

[Myriad]

I don't think the argument here is only, or even mostly, about the definition of "random." It's also rather fundamentally about the definition of "outcome."

The point of my analogies with air pressure, fission chain reactions, and so forth is that whether you see the "outcome" of a set of randomly decided events as "random" depends on what level of description you look at the system. The movement of air molecules in the tire is random, and which molecules collide with the wall at any given time interval will also be random. But the cumulative phenomenon we call gas pressure is not random. Thus, we can call the outcome random if we care whether Aaron the Air Molecule hits the wall during some specific time interval, but if what we care about is that the rims stay off the pavement, it makes no sense to call the overall phenomenon, gas pressure, random.

Same thing for the chain reaction, which is non-independent events. If the 100th nucleus to fission sends a neutron one way, a certain other nucleus (call it Leon) gets hit and fissions immediately. If it sends it a different way, then Leon might never get fissioned at all by the time the chain reaction is complete. So, the outcome is random if we care about what happens to Leon. But overall, the outcome of the chain reaction is not random, it is the same every time.

An even better analogy might be the Barnsley Chaos Game algorithm. Take four affine transformations of a plane (illustrated here.) Start at an arbitrary point on the plane. Then, repeatedly choose one of the four transformations at random, and apply it to the previous point. What you get is the familiar "fern" image, shown here.

Or rather, what you get is a very different set of specific points each time.

There's an uncountably infinite number of points in the attractor, but each running of the algorithm can only collect a countably infinite set of points (and that's only if you run it through infinitely many iterations), so the chance of any specific randomly chosen point being in the set produced by a given run approaches zero. If you don't put any precision limits on your calculations, then each transformation is going to add more nonrepeating decimal digits to the next point's coordinates, which means that any given point in the attractor has at most one chance to be in the set, and only if a unique, or one of a very few, combinations of random choices has been made for a (potentially very large) number of steps.

But, you still get the same picture every time.

Random input to the process. Random outcome, if you care whether Pam, Peter, Paul, Phillip, Polly, or Petunia the Point end up in the set or not. Non-random outcome, if you're looking for a picture of a fern.

So, what do you consider the "outcome" of evolution?

If the outcome is "a large set of specific individual genomes," then yes, the outcome of evolution is random, in the sense of being completely sensitive to and completely controlled by the random inputs from outside the system.

If the outcome is "a world full of diverse and complex life," then no, the outcome is not random. It'll come out that way every time.

Respectfully,
Myriad

 

[Taffer]

The really interesting question, to me, is why is it so important? In addition to spawning threads on JREF, Dawkins spends several pages in The God Delusion on the subject. People seem to have strong opinions.

Again, because it leads to key misunderstandings of evolution.

In the sense used in every math textbook in the world, evolution is very clearly random. In almost every sense of the word found in the dictionary, evolution is clearly random. There is one and only one sense found in the dictionary in which evolution is not random. That's the sense of "all outcomes being equal". That clearly is not the case with evolution, but neither evolutionists nor creationists say that it is. We can disregard that sense because the context makes it clear we are not using that sense.

Perhaps. But how many times have you heard a creationist claim evolution is impossible, because it is just 'random'? How many times have you had to explain that evolution is non-random selection acting on random variation, and that is why it can work? How much time have you spent studying the models which show that random variation alone is not enough to drive evolution?

Usually, JREF is pretty sympathetic to using the technical, mathematically precise, definition of a word. Why is this one different?

As usual, I'll offer my opinion. It is partly because the term has occasionally been misused by creationists. Some of them say "evolution claims that animals are assembled at random." Evolution doesn't say that, and some scientists (both professional and amateur, like us) bristle at the comparison.

I confess I do.

The assembly of organisms follows instructions in the DNA, and the DNA itself, while the result of a random process, is assembled from its predecessor version of DNA, with perhaps a few random errors. No molecule of DNA, except possibly the first, very simple, molecule, has every been assembled randomly.

Correct. Well, essentially.

However, I don't think that's the full explanation. Despite the misuse of the term, the actual term, when used correctly, is very precise, and very accurate in describing evolution. To give my opinion, I'll turn to Dawkins' explanation from "The God Delusion".

Actually, it's from "Climbing Mount Improbable", but I haven't read that book. I've only read the shortened version of the argument in "The God Delusion". In that book, he describes evolution with an analogy of someone staring at a cliff, and saying, "No one could ever go up that cliff!" The "cliff" is the assembly of complex organisms from simple molecules. However, on the other side of the cliff is a gentle slope, that can be walked up easily. The gentle slope is the slow variation of organisms over time. Each step is quite easy, and with time, the "summit", a complex organism, is reached through that series of small steps.

Every analogy has weaknesses, and we can see what Dawkins is getting at here, so the analogy works. However, I think it is not a very good one. I think a better one would be to replace the gentle slope with a pockmarked, canyon strewn wilderness. It is filled with craters, small cliffs, landslides, low summits and deep crevases. Organisms grope blindly in all directions all over the wilderness. Eqrthquakes reshape that wilderness periodically. In the wilderness, though, there is a path leading to the summit. It goes up and down all over the place, but it does get there, and it gets there with no single step that requires a large cliff. Lots of organisms start at the bottom, and their descendants end up in every nook and cranny of the wilderness, including, eventually, one that finds its way to the top, after many, many failures along the way.

I think that's a better analogy, but it isn't very appealing. It makes it seem like the achievement of the summit isn't an "achievement" at all. Eventually, by sheer dumb luck, someone will make it to the top. That isn't as exciting as saying that the complex organisms, i.e. us and our pets, are part of a magnificent adventure that makes clear the amazing power of nature. Nevertheless, it is accurate. We have achieved a temporary summit on the evolutionary hill, and we did it by chance.

Good analogy, and I would agree with it. But you should add more to it. There is a force which pushes those which are unfit towards a more favourable route. If the path taken was completely random, then the chance of getting to the top is tiny. But given this 'force', it becomes far more likely. And if we knew every hill and valley of the wilderness, we could say which path the population would take. In other words, if we knew which mutations would happen, and when, and if we knew the fitnesses of those mutations, we could say how selection would drive the frequency of alleles in the population over time.