In the Roger Scruton on altruism thread, some commenters have expressed confusion over the evolutionary explanation of altruism in ants. If workers and soldiers leave no offspring, then how does their altruistic behavior get selected for?
The answer is simple but somewhat counterintuitive. The genes for altruistic behavior are present in both the workers/soldiers and in their parents. Self-sacrificing behavior in the workers and soldiers is bad for their copies of these genes, but it promotes the survival and proliferation of the copies contained in the queen and in her store of sperm. As long as there is a net reproductive benefit to the genes, such altruistic behaviors can be maintained in the population.
Selfish genes, altruistic individuals.
Let’s dedicate this thread to a discussion of other counterintuitive evolutionary truths. Here are some of my favorites:
1. The classic example of sickle-cell trait in humans. Why is a disease-causing mutation maintained in a human population? Shouldn’t selection eliminate the mutants? Not in this case, because only the unfortunate folks who have two copies of the allele get the disease. People with one copy of the allele don’t get the disease, but they do receive a benefit: improved resistance to malaria. In effect, the people with the disease are paying for the improved health of the people with only one copy of the mutation.
(Kinda makes you wonder why the Designer did it that way, doesn’t it?)
2. In utero cannibalism in sharks:
Shark embryos cannibalize their littermates in the womb, with the largest embryo eating all but one of its siblings.
Now, researchers know why: It’s part of a struggle for paternity in utero, where babies of different fathers compete to be born.
The researchers, who detailed their findings today (April 30) in the journal Biology Letters, analyzed shark embryos found in sand tiger sharks (Carcharias taurus) at various stages of gestation and found that the later in pregnancy, the more likely the remaining shark embryos had just one father.
(Kinda makes you wonder why the Designer did it that way, doesn’t it?)
3. Genetic conflict between parents and offspring. Here’s a great example from a 1993 paper by David Haig:
Pregnancy has commonly been viewed as a cooperative interaction between a mother and her fetus. The effects of natural selection on genes expressed in fetuses, however, may be opposed by the effects of natural selection on genes expressed in mothers. In this sense, a genetic conflict can be said to exist between maternal and fetal genes. Fetal genes will be selected to increase the transfer of nutrients to their fetus, and maternal genes will be selected to limit transfers in excess of some maternal optimum. Thus a process of evolutionary escalation is predicted in which fetal actions are opposed by maternal countermeasures. The phenomenon of genomic imprinting means that a similar conflict exists within fetal cells between genes that are expressed when maternally derived, and genes that are expressed when paternally derived.
(Kinda makes you wonder why the Designer did it that way, doesn’t it?)
Can readers think of other counterintuitive evolutionary truths?
Addendum
4. Mutant organism loses its innate capacity to reproduce and becomes a great evolutionary success. Can anyone guess which organism(s) I’m thinking of?
How does the program know what is better and what is worse? There is your fatal flaw.
Think about it, WJ. If you’ve already determined the fitness function and set a fitness peak as your halting condition, then clearly the parameters for that peak are not the results you’re after, since you already have them. If you write such a program, it’s because you want to know something else.
Where have we heard this before? Oh, I remember!
phoodoo on January 7, 2014 at 4:29 pm said:
🙂
You define a fitness function, ‘doo. An example. You have an environment with agents competing reproductively, and you tell your agents: ‘if you can’t find enough food, you won’t be able to reproduce’. Now the fitness peak is defined as an optimum in food gathering. But what we’re after is the food gathering strategies developed – and those weren’t coded for.
Have you heard the term fitness function? Surely you have.
phoodoo,
The fitness function provides a means of differentiating among strings. Just as being the wrong size, colour, speed or whatever provides a means of differentiating among organisms in the old Struggle for Life. You could easily introduce an environmental determinant to supply the differential to the program, if you so chose. You could even do it blind – the programmer does not need to know what parameters are going to be supplied. It would just be pointless, other than an illustration to the obtuse.
olegt,
Heh heh!
So, evolution works huh? Fitter varients are more likely to stay alive.
Remind me what ID is needed for again? You seem to have forgotten about that.
For 20 years people paid me to write programs, when all the time I, or the person writing the specs, knew the answers they would produce.
Allan,
Evolution doesn’t have a definition of what is fittest, other than what exists. So why doesn’t your program just decide that whatever output you get is what is best?
You are afraid to call Joe out, on his statement that an evolutionary algorithm can show you just how ants derived their social tendencies, but you know full well this is false.
Its every bit as false as Olegts claim that you can kill off M&M’s indiscriminately in a reasonable amount of generations and achieve successful drift of a new allele, without first bastardizing what a generation is. You are not dealing in truth, so your programs also don’t.
What is the fitness function that tells you how ants became altruistic?
Yes it does, although it’s a fairly broad definition: anything that helps a particular variant outreproduce its competitors.
Perhaps he would know that, if it weren’t for the fact that such models actually do exist, and they do work.
Hm?
After all this time, phoodoo, are you still confused about the M&M generation count? 🙂
Not in the slightest.
Gralgrathor,
Name one, that does what evolution does.
Allan Miller said:
Only if evolution is a process that includes deliberately constraining the parameters of “random” variations in service of a goal, and deliberately utilizing a selection process that is the best fit for acquiring that goal, can on say that simulators are doing “exactly” what evolution does.
I doubt you’d agree, however, that this is what evolution is doing.
I have no problem with the characterization of evolutionary processes as one of variation and selection, nor with the comparison of computer simulations to the real world as one of modeling variation and selection. That is what such models actually do.
The problem lies in the qualifying terminology: random variation and natural (as opposed to intentional) selection. These are ideological assertions, not scientific, and is the conceptual error in the claim that the sims model “evolution” in the real world. The computer models represent a controlled parameter of random variations suitable to the goals of the simulation (not truly “random”), and a selection process that gives the best chance of acquiring the target conditions.
The qualifiers “random” and “natural” is what marks the distinction between what is presumed to be occurring in evolution in the real world and what is actually occurring in a simulation.
Gralgrathor said:
Except I didn’t say you set a “fitness peak”; I said you set parameters for what fitness means in terms of a goal.
If you want to find the optimal configuration of various kinds of wooden elements for winged flight, you don’t include rocks, metals and prime numbers in your pool of variants and optimize your selection process to kill off wooden structures that fly, or optimize it to kill off all non-flowering plants. This is all embedded oracle information that is specifically set by so-called evolutionary sims to find a target that matches the goal conditions.
ROFLMAO
Ok phoodoo, with random replication and no mutation:
What is the probability that one of the M&M’s will become the ‘ancestor’ of all the M&M’s in the bag?
What is the probability that THIS M&M (you know, the black one) will become the ‘ancestor’ of all the M&M’s in the bag?
You said you had run simulations and “the results were ludicrous”. A number of posters offered to swap code with you, but you went strangely silent. What programming platform did you use for your simulations?
Presumably you’d consider simulations of parachute behaviour invalid, unless the computer running the simulation was itself thrown out of a plane?
Staying alive. Is that not a goal that has many paths to it?
Do you not know what modelling is then? Are models of the behaviour of gases invalid because there is no smell of gas from the model?
Sure. Tom Ray’s Tierra, an experiment in open ended digital evolution, showed how through differential reproductive success, populations of agents could diversify into prey and parasites.
Tom was not the first to model differential reproductive success using computers: such experiments have been done since the 60s, eg. Interdemic Selection and the Evolution of Altruism: A Computer Simulation Study, Levin et al, 1974; Organizational Evolution, Learning, and Selection: A Genetic-Algorithm-Based Model, Bruderer et al, 1996; The Evolution of Trust and Cooperation between Strangers: A Computational Model, Macy et al, 1998, etc.
People can use and have used computer modelling for anything ranging from finding optimum solutions for a given definition of fitness to confirming evolutionary trends in animal behaviour. There are easily thousands of such papers detailing digital experimental evolution to answer specific questions.
A target, yes. Often, such simulations are run to find such targets – which then can be used as engineering solutions, for instance. At other times, the products of such evolution are of secondary importance, and the experiment is really about testing mathematical models for how particular fitness peaks can evolve.
Oh, let’s check! Here is a simple question, phoodoo.
I have a bag of N M&Ms. N is a large number, let’s say a thousand. I open the bag, eat one M&M and replace it with another, from a different bag.
How many M&Ms do I need to eat (and replace) to call it one generation?
Gralgrathor said:
A target within a pre-determined set of goal conditions. That’s why you don’t kill off those simulated organisms which are better suited to the set goal condition, and don’t include variants in the pool that have nothing to do with acquiring targets that will meet goal conditions.
IOW, the only “random” variation and “natural” selection going on is that which has been chosen and tuned to be best suited to reaching the goal conditions in mind. So, it’s really not “random” or “natural” at all. It is, however, variation and selection.
That’s right! Congratulations. Now, perhaps you can go and explain that at UD?
How does your conceptual model cope with the idea that instead of being pre-determined the goal conditions evolve independently of the organism. That change things at all?
What’s your point anyway William? That you can’t model something and call it valid unless it’s done at the molecular level?
For a real living organism, what’s doing the “choosing and tuning”?
Thanks. I think Sophisticat (whom I hope you haven’t driven away), made some excellent remarks on that subject. This–which reminds me of the writings of the *great* Amie Thommason–was particularly good, I think:
Tierra:
“According to Thomas S. Ray and others, this may allow for more “open-ended” evolution, in which the dynamics of the feedback between evolutionary and ecological processes can itself change over time (see evolvability), although this claim has not been realized – like other digital evolution systems, it eventually reaches a point where novelty ceases to be created, and the system at large begins either looping or ceases to ‘evolve’…
Mark Bedau and Norman Packard developed a statistical method of classifying evolutionary systems and in 1997, Bedau et al. applied these statistics to Evita, an Artificial life model similar to Tierra and Avida, but with limited organism interaction and no parasitism, and concluded that Tierra-like systems DO NOT EXHIBIT THE OPEN ENDED EVOLUTIONARY SIGNATURES OF NATURALLY EVOLVING SYSTEMS…
Russell K. Standish has measured the informational complexity of Tierran ‘organisms’, and has similarly NOT OBSERVED COMPLEXITY GROWTH in Tierran evolution.”
Tell me how again this is going to explain ant behavior?
Remind me how ID explains ant behaviour?
You mean your answer is, we can’t but neither can you? That’s your “scientific” theory?
Hardly. But it’s very easy to criticize, somewhat harder to be constructive.
So please, do explain how ID is a better explanation then that which you think is being offered for ant behaviour.
walto,
Have you considered doing an OP on Thomasson? You’ve mentioned her paper on nonreductive physicalism a couple of times.
NOTHING is being offered as an explanation for ant behavior, so surely its being constructive to point this out.
Then the floor is open for ID to step in and fill that gap, no?
please do so.
phoodoo,
You’re missing a couple of points here.
1. Computer simulations are necessarily limited in scope. To get the kind of open-endedness you find in nature, you’d have to provide enough computing power to model every last elementary particle in an entire ecosphere. But, and take notice, even though Tierra has the same limitations as any computer model today has, it does produce novelty. More computing power will mean more permutations, and therefore models become possible that will approach physical nature more closely and thus produce more novelty.
2. We see novelty being produced in nature. There’s no question that it happens, we just want more details on how it happens. Simplified models of evolution, no matter how limited they are, can help in examining principles and mechanisms. You don’t have to build an actual solar system to project the course of a spaceship either. A simplified model will do just fine.
3. I don’t know what Standish analyzed, but he could hardly have missed the fact that some Tierran populations speciated and ultimately evolved into species of parasites and prey. So even at the coarsest level, we already see an increase in complexity. So I’m inclined to think that Standish carefully chose his methods of analysis to work towards a desired result.
William J. Murray,
The variation in nature is random, and so is that in most computer simulations. ‘Random’ being ‘stochastic’, unless you’d prefer to argue for another definition. But I can’t see a definition myself that would render the processes different with respect to it as regards the generation of variation. EAs are deliberately made as ‘evolution-like’ as possible. That’s the whole point of the method: it has been found to be a powerful approach to certain problems.
Of course the selection is not ‘natural’. Nor did I even use the word.
Sigh.
Gralgrathor,
Given the following abstract from Bedau and Standish, I suspect that RKS has been quote-mined…
Allan Miller said:
Asserting it isn’t demonstrating it.
Within parameters specifically attuned to the target condition.
No, they are made as “evolution-like” as is possible within the framework of acquiring specified target conditions, something that is precisely not “evolution-like”, at least not by the standard meaning of the term.
The problem is that in order for the method to be of any value whatsoever, the pool from which random variants are drawn and employed must be tuned to the target conditions, and the selection process must also be tuned to the target conditions. You don’t get squat in the lab or in the computer sim without the tuning the system to explore towards the specified target conditions. Like altruistic behavior. Or winged flight.
William,
In your opinion, is there such a thing as a “fair die”?
Given those tuned “parameters” how many states are nonetheless available?
Is “staying alive” or “reproducing” not also a target condition?
I see. What did/does that tuning in nature would you say?
Oh? In the “robots evolve altruism” example, what was tuned?
http://evolution.berkeley.edu/evolibrary/news/110501_robots
And going back to the first point, what specifically makes you think that variation in nature is not random? Anything specific?
So, you are saying that unless there is a designer doing some tuning, evolution does not and cannot work and therefore your “evidence” for ID is that as evolution cannot do the things claimed for it and so ID is actually the answer?
Am I understanding you correctly?
Or are you just making the obvious point that simulations are, by their nature, incomplete? If so, no need to bother, you are not saying anything new.
I could be wrong, but I was under the impression that “random variation” as it’s used in evolution talk, doesn’t mean “anything goes.” I mean, my wife and I couldn’t have had a lemur for a child, or a kid with wings or a child that could make himself invisible. If I recall a popular Dawkins book on this matter correctly (and if he was right), there are limits to the available mutation possibilities in nature. That being the case, there is no problem with computer simulations that also have limited possibilities. All that’s necessary is that the probabilities aren’t jiggered. And given the combo of natural selection and a ton of time, it’s my generally uninformed sense of the matter that you can get pretty impressive results–not just lemurs and animals with wings, but human beings–without violating the premise that mutations are random.
Computers are great for that kind of demonstration, it seems to me.
People like phoodoo and WJM seem to be calling for completely realistic simulations of reality. Now if this were physics instead of biology that would be like saying that since we can’t do a completely realistic simulation of the baking of a cake, that therefore physics doesn’t work.
In the present thread the issue was altruistic behavior of social insect castes. The question was whether, when a behavior has a negative impact on the individual’s fitness, but a positive impact on the fitness of some relatives, a gene predisposing to this behavior can have its gene frequency increased by selection. The answer is yes, under the right conditions. Those conditions are given by W.D. Hamilton’s famous 1964 inequality. And you can do computer simulations verifying this math.
Anyone who holds out for a fully-realistic model of evolution before they are willing to see the logic of a kin-selection (or group-selection) argument is just being silly. As if the field of physics were waiting with bated breath for a realistic simulation of baking a cake.
I think it’s more about “designer in the gaps”. If I can’t prove that something is not interfering in what appears to be a random process that’s actually evidence for ID!
I know, I know, but yes, really. I’ve been asking the “Can a die be random” question for some time now, no ID supporter seems to want to answer it for some reason….
Indeed. If only there was a way to programatically measure FSCO/I (or whatever) then we could feed the output of various existing simulations into it and see if FSCO/I (or whatever) is actually being generated or not.
If it is, well, no designer required to generate all that FSCO/I nonsense the IDers observe (but can’t quantify for some odd reason) in biology. If not, well, score one for the ID crowd. I’d be happy to give it to them too, they would have dun a science at that point!