Randomness in Evolution: Valid and Invalid Usage

[Herzblut]

here the random walk variable is X, and since it has a memory it is arguably already a case of "biased randomness", but the steps in the random walk (the Z's on this page) are IID (that is "pure random"). as you might guess, most of the 407 other pages in this book focus on more interesting varieties of "biased randomness" - referred collectively as "stochastic processes".

This simple random walk has interesting features already.

For instance, the random walker will find himself almost always on the right-hand-side (X>0) or almost always on the left-hand-side (X<0).

The case where he finds himself more or less equally often left or right, is the least likely of all cases.

 

[jimbob]

thanks Reality Check

The confusion comes because certain posters are trying to apply their interpretation of the word random (which seems to be "non-deterministic") to the entire theory of evolution.

yes, i see this. i introduced the "random walk on a tilted surface" as a step toward stocahstic motion on a fitness landscape, naively assuming both sides would agree.

but i see why, for tactical reasons, one side goes quiet.

this may be an issue in public presentation of natural selection, but it is not an issue in statistics or in physics (or amongst researchers in natural selection speaking in private). from a pedagogical point of view it seems a high risk strategy not to simply explain that stochastic dynamics does not mean "utterly random".

but i have no desire to add fuel to the confusionist argument.

I would say that evolution is not random but it is also not deterministic.

i understand. and i may understand why you would say that. nevertheless mathematically (and philosophically) a process is either determinsitic or not determinisitic (sets of measure zero excluded), and eventually i expect biologists will have to declare. i personally believe declaring early is better, but you guys are the ones fighting the fight. i hope understanding prevails in the end!

thanks for your reply.

I'd prefer "random" but not haphazard". For many situations, I would have thought that the analogy of a bent game of dice where the randomness affects the score, but not who wins.

Furthermore, Lenny, in my experience in the UK, I mostly (although not always) come across religious people who "accept evolution" but think that it is somehow predetermined and that God could have set the ball rolling, with the inevitable consequence of something akin to humanity arising.

This is not what we see evidence for at all. Obviously it was not impossible, but there were a myriad of (almost unimagnible) other possible ecosystems, it was merely chance that led to the current ecosystem...

On the issue of strategy - that's largely a question of how best to communicate the concepts underlying evolution. I think the answer is to explain that there are random elements, like a large meteor hitting earth, and that there are non-random elements, like the effects on species of the climate resulting from that meteor strike.

In the case of those bacteria, which petri dish first evolved the ability to metabolize citrate was totally unpredictable, but the fact that that particular ability might evolve (out of an infinity of possibilities) was predictable.

I don't think anyone above chimpanzee level will have trouble understanding that a process can have both random and non-random elements. But when you say "evolution is random", you are communicating to most people the claim that there are no non-random elements, which is simply false.

I differ with this analysis of the long term evolution experiment, and the evolution of citrate metabolism.

Especially:

but the fact that that particular ability might evolve (out of an infinity of possibilities) was predictable.

Adaptations to environments will occur; what form these adaptations will take is affected by historical contingency and the history of the previously "neutatral" mutations.

With a more complex original ecosystem replicated in different expeiments, the effect of a chance mutation affecting the evolutionary path of the other types of organisms would be amplified.

If a citrate-metabolising bacterium was introduced to these populations around generation 30,000 in some flasks it might be able to establish itself, having no competition, whilst in others it might not, depending on the effeciancy of citrate metabolism of the established e.coli populations.

As soon as you introduce different competing types of organisms into the mix, the whols fitnes landscape is subject to random alteration when "disruptive" mutations occur. This change is also random, as well as the actual mutations themselves.

 

[mijopaalmc]

Come see articulett embarrass herself when she discusses the stochasticity of evolution with real evolutionary biologists.

 

[Reality Check]

It looks like you win: Evolution is random.
So what?

 

[mijopaalmc]

It looks like you win: Evolution is random.
So what?

If you are agreeing (rather than conceding to stop the argument), why is there such antipathy to a simple statement of fact?

It is as if I had "evolution is an observable fact" to a creationist.

 

[Reality Check]

If you are agreeing (rather than conceding to stop the argument), why is there such antipathy to a simple statement of fact?

It is as if I had "evolution is an observable fact" to a creationist.

The problem is that you did not define what you mean by random. So everyone thought that you mean random as in unpredictable. Now everyone knows that you (probably) mean random as in a stochastic process (i.e. a process with predictable outcomes).
So evolution is a random, predictable process.

 

[mijopaalmc]

The problem is that you did not define what you mean by random. So everyone thought that you mean random as in unpredictable. Now everyone knows that you (probably) mean random as in a stochastic process (i.e. a process with predictable outcomes).
So evolution is a random, predictable process.

Actually I did define "random" from the outset of the discussion and have repeated that definition several time of its course, but people rejected it saying that it "makes everything 'random'" and fell back upon the non-technical definition.

 

[Reality Check]

Actually I did define "random" from the outset of the discussion and have repeated that definition several time of its course, but people rejected it saying that it "makes everything 'random'" and fell back upon the non-technical definition.

From memory, that was a result of you using non-deterministic as part of the definition of random. So that makes a lot random (not everything) including the example of a smoke detector (IMHO not a good example).

 

[mijopaalmc]

From memory, that was a result of you using non-deterministic as part of the definition of random. So that makes a lot random (not everything) including the example of a smoke detector (IMHO not a good example).

But scientists do often use "random" to mean "non-deterministic". In fact, the deterministic/non-deterministic distinction often makes much more sense for the usage of "random" than the uniformly distributed/non-uniformly distributed distinction. This is because any non-uniform probability distribution can be transformed in the appropriate uniform probability distribution, whic means the same thing is both random and non-random, if "random" means "non-uniformly distributed".

 

[Reality Check]

But scientists do often use "random" to mean "non-deterministic". In fact, the deterministic/non-deterministic distinction often makes much more sense for the usage of "random" than the uniformly distributed/non-uniformly distributed distinction. This is because any non-uniform probability distribution can be transformed in the appropriate uniform probability distribution, whic means the same thing is both random and non-random, if "random" means "non-uniformly distributed".

That is correct. Using non-deterministic for the random parts of the process of evolution is also correct. Using non-deterministic (or random) for the entire process of evolution is confusing since there are deterministic parts within the process (natural selection).

 

[mijopaalmc]

That is correct. Using non-deterministic for the random parts of the process of evolution is also correct. Using non-deterministic (or random) for the entire process of evolution is confusing since there are deterministic parts within the process (natural selection).

You last statement is incorrect. If any part of the process is random, the process is random as a whole. By the way, the vasat majority of the discussion seems to be predicated on the idea that mutation is random and natural selection is not while presenting very little evidence that natural selection is not in fact random and seemingly ignoring the fact that the empirical evidence (i.e., Lenski's Long-Term Evolution Experiment) contradict that assumption.

 

[Reality Check]

You last statement is incorrect. If any part of the process is random, the process is random as a whole. By the way, the vasat majority of the discussion seems to be predicated on the idea that mutation is random and natural selection is not while presenting very little evidence that natural selection is not in fact random and seemingly ignoring the fact that the empirical evidence (i.e., Lenski's Long-Term Evolution Experiment) contradict that assumption.

Your second statement is incorrect. If a random process is followed by a deterministic process then the process as a whole is not random. For example throwing a die is random but then selecting a specific number from the throws makes it deterministic (whatever you throw you always get that number).
Evolution is the highly random process of mutation followed by the less random process of natural selection. The overall process is somewhere between random and deterministic. The usual term used for such processes is stochastic.
The empirical evidence is that evoloution is stochastic (random by your definition), predictable and repeatable (as in Lenski's Long-Term Evolution Experiment).

 

[sol invictus]

Actually I did define "random" from the outset of the discussion and have repeated that definition several time of its course, but people rejected it saying that it "makes everything 'random'" and fell back upon the non-technical definition.

And those people were absolutely correct. According to your silly definition, either everything in the world is random, or everything in the world is not (depending on something we don't know now and will probably never know, namely the fundamental nature of quantum mechanics).

That makes it totally useless, and it renders this whole idiotic rotting zombie discussion even more pointless, absurd, and stultifyingly ridiculous than it otherwise would have been.

Everyone that participated in this thread had had their IQ permanently reduced by at least 5 points. Congratulations.

All I can say is, "Wake up, you cardboard!"

 

[jimbob]

What about defining something as random if random factors cause a significant (to be defined) alteration to the state of the system?

In this case I would argue that evolution wouold still be random, because if the tape was rewound, and evolution replayed, there is no reason to suppose that similar organisms would re-evolve. At the time of the KT impact, there was no inevitiability in the eventual evolution of humanit, nor of any similar organism occupying similar niches. Gibven the effect that organisms have on the ecosystems, there was not even any inevitibility in the general "shape" of the ecosystem 65-million years later. Although several niches whould have been highly probable.

 

[sol invictus]

What about defining something as random if random factors cause a significant (to be defined) alteration to the state of the system?

Defining terms in terms of themselves isn't usually helpful.

 

[jimbob]

It depends on your point of view. An analogy:

In the second world war balistic missiles were aimed at London.

It was predictible that the missiles would land in London, it wasn't predictible which particular houses would get hit. From the point of view of someone interested in the number of hits on London it was predictible. Form the viewpoint of a householder, it was random whether the missiles would hit their particular house.

Brownian motion might have an infinietessimal effect on the tragectory, but this would not ususally be considered significant.

there wouldn't be a hard and fast rule, but if random factors caused 20% uncertainty, they would be more significant than if they had caused 5% uncertainty. Above 50% uncertainty due to random factors, and I think it is always acceptableto call the system "random".

Supposing you released a ballon into the atmosphere. And at the end of each day, you recorded its position. With a knowledge of weather systems its initial behaviour would be predictible, but it would be impossible to predict where the balloon would be in a year's time*, because random factors would have played a significant part.



*It's a magic balloon that doesn't deflate, and has neutral buoyancy in air.

 

[Walter Wayne]

Since this thread is on the front page, I thought I would mention that the new Skeptic (Vol 14, No 2), might be of some interest to people here. A couple of articles are loosely related to some things brought up in this thread.

Predicting Evolution
How Likely is it that Human-level Intelligence will Evolve Again?

by David Zeigler



The Chain of Accidents & the Rule of Law
The Role of Contingency & Necessity in the Evolution of Higher Intelligence

by Michael Shermer
​

Walt

P.S. I haven't read the articles, just thought I would pass it on.

 

[sol invictus]

It depends on your point of view.

The point seems to have gone whizzing over your head. You defined "random" in terms of "random". That's really not useful.

 

[jimbob]

No, I defined a random system as one where random differences played a significant part in determining the state of the system. With a smoke detector, the radioactive decay might be random, but because of the rate of decay events, random differences do not cause a significant part in determing the state of the system. You can average these out.

It is my contention that you can't with evolution.

If a cosmic ray can cause a seeding event that is of a scale to affect cloud formation, and that this becomes "significant" for a particular defintion of "significant", then weather would be random over timescales greater than this. Especially if a cosmic ray a few millimetres/miliseconds away would cause a "significantly" different outome.

I thought we had already decided what random was.

I'm happy with quantum uncertainty and anything "significantly" affected by that as being random.

My contention is that quantum decay events are random and they can affect the survival of individual organisms, both directly, and (more importantly) indirectly, therefore natural selection is "probabilistic" (a skewed game of chance) and evolution itself*, by the multiple feedbacks and simple malthusian reasoning (most "beneficial" mutations will also fail) is also random although often predictible for long timescales.


*Every so often the fitness landscape will be changed by e.g. a disruptive mutation that alters the rules for many organisms, for example.

 

[Wowbagger]

My monster! It's still growing! Ahhhhhhhhhh!!!!!!!!!!