Another excellent post.

If you don't mind me asking, roughly how long did it take to write, how long do you think it would have taken a year ago, and (assuming reading is already fast) what do you think are the most important factors that make writing speed go voom with frequent practice?

Am I right in thinking that you've now brought the OB audience to where you need them in order to start trying to talk about AI (or "optimizing processes" or whatever terminology is sufficiently abstract to prevent linguistically inferred misunderstanding)?

Wrong. Sometimes quality, not quantity, matters.

Well, actually, it seems to be a case where a simple "ceteris paribus" would have taken care of that. If someone writes something that is easily fixed with a "ceteris paribus" or by some other simple means, I try to avoid saying "wrong".

My priors are not what they were a week ago. Thank you for the fascinating posts.

But the error isn't fixable simply by the addition of "all else being equal", because the important concepts that need to be addressed include the reality that some things are grossly unlikely to ever be equal.

There's a reason humans usually have only one child gestating at a time, even though it would be a simple matter biologically to have multiple fertilizations.

There's another reason the mouse gene that causes 90% of mouse sperm to carry a copy of itself in the heterozygous state and certain fetal/infant death in the homozygous doesn't cause ... (read more)

Caledonian, delaying reproduction in order to maximize an individual organism's lifetime output is not the same as foregoing reproduction to benefit the chances of other organisms. The latter behavior is altruistic, the former behavior is not. Williams discusses this while showing that observed cases of apparent reproductive restraint match very finely the requirements of individual optimization.

Now I've said before that I can't include all the fillips and caveats in a series of blog posts. If you want to add a fillip in a comment, that's fine, but please don't assume that I'm ignorant.

I suggest reading the special issue of the Journal of Economic Behavior and Organization, January 2004, where the matter of multi-level evolution is dealt with at length, with some of the commentators evolutionary geneticists, mathematical ones at that, and at the highest level. The conditions for higher level selection are laid out there. Wynne-Edwards did not know these and was accurately put down by Williams in 1966. They are the Crow-Hamilton-Price equations. I suggest you read it, Eliezer, and, yes, I am the editor of the journal, the leading one in the world on evolutionary economics.

I thought I made it clear that I knew about the modern revival in multilevel selection. Was this not clear enough?

Group selectionism would later revive, somewhat, in drastically different form - mathematically speaking, there is neighborhood structure, which implies nonzero group selection pressure not necessarily capable of overcoming countervailing individual selection pressure, and if you don't take it into account your math will be wrong, full stop. And evolved enforcement mechanisms (not originally postulated) change the game entirely.

Barkley: got any explanations for why the supportable information in a genome should go as the inverse square of the mutation rate?

"Rationalisation of a predetermined bottom-line" is not always be a bad thing. It is common enough in Mathematics that you intuitively feel a result is right, and you work backwards from the result to see how you can prove it. The real mistake is if you do not take care in working it out backwards, and make wrong inferential steps in the chain. You may (legitimately) point out failures of this strategy, but there are also successes that you need to acknowledge.

Eliezer: It's still not at all unusual to see serious biologists advocating magical thinking. Evolutionary theory as conveyed in the undergraduate biology curriculum is extremely elementary, and it's easy to graduate without even mastering what is covered, not to mention without making single step inferences from it. Most biologists know no more about modern evolutionary than that plus the results of reading and for the most part believing either Gould or Dawkins. You can't assume that scientists are familiar with the sub-fields of their discipline, no... (read more)

Venu: You definitely can do that in Mathematics, but that's because reasoning about Mathematics has some special properties that most reasoning about the real world does not.

1. Math is black and white If you find a proof for something, it's true. Until you do, you can't really call your hunch math. However, in the real world, it's very easy to find arguments for things that are false.
2. Math has monotonicity What this means is, that if you use Lemma A and Lemma B to prove Theorem Z, then whether or not Lemma C is true has nothing to do with whether your proo

Eliezer,

Maybe I will get back to you on that point, and maybe I will not. I am about to leave town, and I spend 30 hours a week editing a journal, along with being a full professor of economics. This is why I turned down Robin's invitation to become a co-blogger here. Too damned busy. This is not meant to be a copout or an escape. I already spent a couple of hours I did not have last night digging through Gould again on your other posting.

I will note that the special issue I edited includes a wide variety of views and that there remain sharply contras... (read more)

Wiseman, you are not describing group selection. You are still describing individual selection, because the causally effective advantage is to the individual. The benefit to the group is a side-effect. Here is your description of the advantage, which you call the key: "because it will spend less energy developing fox fetuses that won't survive anyway." That is an advantage to the individual foxes.

I also have doubts about the specifics of your scenario, but I won't get into that.

Constant, It's group selection because the individual is essentially making a sacrifice to reproduce less, to benefit the group. It happens blindly, through normal evolution of selecting the individual, but how else do you expect it to happen?

Eliezer -

Would "innovation" in genetic error correction, or changes to the proteins responsible for allowing greater or fewer mutations in DNA...

...would such "meta-changes" (changes to the mechanisms of DNA replication) be the basis for group selection?

If different groups had slightly different rules for their DNA replication, intuitively I could see that their competition would be best understood as group selection.

Consider two groups, both formed by mating of a single mother pregnant with a son, leading to two groups with slightly di... (read more)

Constant, It's group selection because the individual is essentially making a sacrifice to reproduce less, to benefit the group. It happens blindly, through normal evolution of selecting the individual, but how else do you expect it to happen?

As I pointed out, the benefit to the group is a side-effect. In your scenario the fox survives because of a direct benefit to the fox. As for "how else" "I" expect it to happen, it's not about what I expect (since I am not advocating group selection), it's about what group selection is. As Wikipedi... (read more)

Wiseman, you need to put your scenario into mathematical terms, or write a simulation, or something. It's too easy to imagine some foxes and rabbits breeding and scurrying about, and convince yourself that something is possible. In any case the situation you described is not "group selection", but good old-fashioned gene-level selection. In this case it's selection for genes that lead to an optimal breeding rate.

(Oops, I didn't refresh for a while and I see you beat me to the critique, Constant.)

Constant,

Believe me, I fully see the obvious, but false, contradiction that you point out. Please understand I considered that when I first wrote my example.

It is ONLY a benefit to the individual because it's also a benefit to the group. Under ANY OTHER circumstances, a fox would do better for itself, and only itself, to reproduce more. But because the other foxes, the group, are around, the individual fox has to evolve for selection pressure not just from the non-fox enviroment, but the fox-group enviroment.

The benefit to the group is not a side effect, i... (read more)

Wiseman, I still disagree but am not going to pursue it.

This may be a stupid question but... what about kin selection? How did that develop? Wouldn't something like group selection have had to have happened at some point for kin selection to end up showing up in the first place?

ie, imagine a couple families/clans/whatever of some species. one happens to have a member that has "magic gene(s) of kin selection juju", and the other... doesn't.

Let's say eventually in the former, sometimes members having that gene get to breed just often enough so that the gene/complex/whatever starts spreading around throu... (read more)

Wiseman, you haven't shown that it really is beneficial to reproduce less in the scenario that you are describing. Yes, a smaller group will consume less food - but if there are six foxes, then the probability than at least one of them will survive can very well be higher than if there is only one. The group that reproduces less will still be outbred by the one that reproduces more, so the faster-breeding one could on average have more surviving members.

This might not be the case in situations where food is extremly scarce, but it should be so in situation... (read more)

Kaj, fast breeding does not just incur a cost on the cubs, but on the mothers developing the cub fetuses. No matter the dearth of rabbits/food, as long as it's less than the amount needed to sustain the current fox population, the less energy and time spent by a fox mother developing unnecessary fetuses, the less likely she will die before child birth. You can't just calculate the raw probability of cubs surviving by saying "Each cub has X% chance of surviving, therefore the more cubs, the greater total chance that some will survive". A cub is ta... (read more)

Psy-Kosh, kin selection means that you help those who are closely related to you, which helps spread your genes since those who are related to you are more likely to carry your genes. It's beneficial to breed, since (in sexual reproduction) your children share 50% of your genes. It's likewise beneficial to help your siblings, since they, too, share 50% of your genes. Hamilton's rule states that an allele for altruistic behavior will spread if the behavior it causes obeys the equation Br > C, where r is the relatedness between the actor and recipient, B ... (read more)

Kaj: Ah, thanks. Then I guess I was a bit unclear as to what counted as group selection. ie, I thought a family would count as a "group" for these purposes.

I'll try to quickly program a rough model for it to see what sort of numbers it produces (though I'm no evolutionary biologist).

Never mind. I tried it, but realized that I had to resort too much to guesswork for it to be of any use. (Plus it's getting too late for me to really think about the model.)

Ok Kaj, I agree fast-breeders will at some points overwhelm slow/restrained breeders, at times where food is plentiful and greater than the amount needed to sustain the current fox population. But as long as that breeding goes unrestrained, the ecosystem enters a state which there exists less fox food than needed. As soon as that happens, restrained breeders have an inherint advantage because they waste less energy developing innevitably unviable fetuses. The important thing about this rule is it applies to any situation where they is less food than needed... (read more)

Wiseman, you are not giving an example of group selection. You are imagining one single group (or two if we split the rabbits and foxes apart) over a long period of time. With group selection there are multiple groups, some of which die out on their own or get squashed/absorbed by other groups or the groups increase in size and split apart at different rates. For the rabbits/foxes example we could imagine multiple populations all separated and say that in all instances where breeding was not restrained, they overpopulated and died out, leaving only the ones that did restrain. However, all those populations would be vulnerable to the overbreeding mutation suddenly appearing, so it would not be a good explanation.

TGGP, your description of what group selection is is not in contradiction with mine. I merely described one isolated group, but the concept can apply to more than one of course. Imagine two groups of foxes and rabbits, one in which restraint is developed and selected for because of the greater health of their youth in times of famine, and one in which restraint is not in any gene, in which case the health of that population is generally lower than the restrained group, but still alive because it is not competing with any internal restrained-breeding indivi... (read more)

Barkley: The mathematical population geneticists were defending multi-level evolution, at least in principle under the right conditions.

How on earth could a mathematical population geneticist do anything else? In principle under the right conditions, you can create group selection in a laboratory and observe the results. Price's Equation in its various forms is a logical tautology. The question is whether the tautology has nontrivial empirical content: are group selection pressures tiny trivial things easily overwhelmed by individual selection pressure... (read more)

Wiseman, you are not describing group selection. You are still describing individual selection, because the causally effective advantage is to the individual.

So repeat the logic, only substitute small, interrelated, relatively isolated groups of foxes for the individuals in the previous examples.

Populations can be the unit of selection just as individual organisms can.

Eliezer, I don't need to make assumptions about what you do and do not know when you make statements like "Obviously, selection on the level of the individual won't produce individual ... (read more)

Dude, trying to gain an evolutionary advantage by "restraint in breeding" is like trying to get rich by overpaying your income tax. Hey, every extra dollar you give the IRS is one dollar removed from the national debt, which should improve the economy... and give you as an individual some utterly microscopic imperceptible benefit, w00t!

Dude, trying to gain an evolutionary advantage by "restraint in breeding" is like trying to get rich by overpaying your income tax.

No, it's more like putting your money into prudent long-term investments instead of buying lottery tickets. Trying to pump out as many offspring as possible is an extraordinarily bad strategy, which is why so few species ever attempt it. Even the ones that starve to death while guarding their tens of thousands of fertilized eggs don't try it.

Everything Wiseman is describing is happening at the level of the gene, not the population.

Imagine there is a gene for breeding rate - different variants of the gene give rise to different breeding rates (1, 2, .... offspring per year, let's say). A fox that has a high-rate allele of the gene will spend more energy on breeding than on caring for existing offspring, while the reverse is true with a fox that has a low-rate allele.

Given the natural fluctuations of food availability over the long term, there is going to be an optimal range of breeding rates. ... (read more)

You can alter the question slightly to permit a very limited form of group selection - you have to have completely isolated genomes, to start with, and a high level of mutative cost between the two groups. (I/e, mammalian versus octopus eyes - refinement guarantees the two groups can't crossover or mutate to adapt the other's characteristics.) If selective pressure favours one of the two characteristics, one group will be effectively "selected out."

The genetic variance doesn't even have to be defined - it could just be a selective tendency agai... (read more)

Here are some Gene Expression posts on group selection. It is of course as Eliezer said, an empirical matter, but given the variation within populations compared to that between populations, it doesn't seem especially likely.

Group Selection (oh no not again!) Defining Group Selection: Price's Equation Group selection, the paramters Group Selection can work, just... (this features a model from Henry Harpending that would result in group selection, but the assumptions made may not be probable)

I've heard that E.O. Wilson is trying to wreck the consensus on gr... (read more)

Would the observed instances of Lamarkian inheritance change the debate re: group selection? http://www.newstarget.com/020068.html

To get group selection out of this scenario, you would have to have one fox group with a lower-than-optimal breeding rate, which let the rabbit population expand, which lessened the chance of a crash in food supply that would wipe out the population. Then that fox group would survive, and the neighboring groups would perish. But there is no way to enforce this pact of lower-than-optimal breeding rates in the first place.

Enforce? You don't need to enforce something that's build into organism's biology - and in the scenario you describe, their reduced rat... (read more)

Rosser: I am not going to get into the debate about the specific math of your model as others have already done so.

There's only one equation in this post and it's a standard one. Were you referring to the Speed Limit post? I already put all original math in that post into abeyance pending further investigation.

I'm reasonably certain the McCabe I was quoting isn't your McCabe.

I suppose I have a unique professional perspective on how tragic the class of warped thinking revealed in old-style group selectionism is likely to be. ("You can get all these... (read more)

"Once you are dealing with hominids, which may be the most important example, indeed "enforcement" may well be important. There is a growing lit on how reciprocal altruism ultimately depends on punishment of free riders, that is, enforcement."

That sounds to me like an example of an "Evolution Fairy".

Eliezer,

So, who is your McCabe?

Wiseman,

Wise crack.

Once you are dealing with hominids, which may be the most important example, indeed "enforcement" may well be important. There is a growing lit on how reciprocal altruism ultimately depends on punishment of free riders, that is, enforcement.

Really? I'd thought that was generally understood - that was the whole point of Tit for Tat, after all, that it could both reward cooperative behavior and punish defection. One without the other is useless: kindness without cruelty is weak, cruelty without kindness self-destructs.

But the point I'm trying ... (read more)

This might be a bit off-topic but I have a post that discusses misunderstanding Darwinism and how irrelevant what pleases us is to the evolutionary process, the distinction between natural and artificial selection and the issue of social interaction. It's mostly about that weird kid who shot up his school in Finland though.

Glad you liked the links Barkley. I wish I could have found the link on Trivers I was half-recalling though. Here is a link on E. O. Wilson and group selection.

Caledonian,

Your distinction between altruism and the more general "group-friendly" is useful and relevant.

TGGP,

Regarding artificial selection it is worth remembering that this was one of the major examples that Darwin himself emphasized in Origin of the Species, the efficacy and effect of efforts at artificial breeding and selection by humans of both plants and animals.

In the brief period of time since I last commented two new good posts on group selection, E.O. Wilson and how it does not gel with our aesthetic/moral preferences have appeared.

'Would "innovation" in genetic error correction, or changes to the proteins responsible for allowing greater or fewer mutations in DNA...'

'...would such "meta-changes" (changes to the mechanisms of DNA replication) be the basis for group selection?'

If they can't interbreed, then you get selection like that between two different clones of bacteria. Either the better species survives, or they both survive in their own ecological niches.

If they can interbreed then you might get evolution by group selection but it isn't the way to bet. You'... (read more)

TGGP, Steve Sailer's post seems to me to confuse executing adaptations versus maximizing fitness: "The good news is that conquering land really doesn't pay these days, so peace has become, from a group-selectionist point of view, more rational than in the past." So what? Modern humans will go on executing the conquest adaptation.

I wish I could have found the link on Trivers I was half-recalling though.

Could it be something like this exchange between Trivers and the authors of "Unto Others". Some of the comments from Trivers are priceless.

Caledonian is right, it's not a "fillip" to point out that reproductive restraint can evolve by individual selection in the sense of K-selection, not in the sense of altruism. To be fair, several recent articles in favor of group selection also talk about "reproductive restraint" and mean altruism, but that doesn't IMO excuse it. Any adaptation leading to late reproduction and less offspring in, say, elephants, must have looked like "reproductive restraint" at some point (although not necessarily at the level of numbers of gra... (read more)

Sadly, I do not have a reference currently, or I would have put the scientific name of the gene variant in question.

What I recall - which may or may not be accurate - is that the gene has a minor negative effect on the overall reproduction of heterozygous males, but not enough to matter on the short scale during which it spreads through the population.

Windy, group selection has been hypothesized to be necessary or sufficient in this case almost from the time the gene was first observed.

Presumably other genes could counteract its effects for an alternative method to prevent extinction. Such genes would be weakly selected while the t allele was rare and very strongly selected when it was common. Strong founder effect.

theoretical study:
https://lirias.kuleuven.be/handle/123456789/41135

I too remember a description of a small reduction in survival of heterozygous individuals, which doesn't nearly balance the increase in survival of the gene. And I too don't have a link handy for that.

T gene is one scientific name of the variant - or rather "t haplotype". Geneticists are not as anal about names as taxonomists...

Wiseman's misunderstanding of group selection demonstrates why this would have been an important distinction to make.

Windy, the point you referred to from Caledonian is not different than my own, so clearly it is you who is misunderstanding something here.

I "get" what group selection is, as you know, at the high level it's not a difficult concept. But my point in an earlier argument is that the idea of group selection can logically only mean one thing, and it is not the idea that somehow the group can flourish while the individuals are slowly dyi... (read more)

Windy, the point you referred to from Caledonian is not different than my own, so clearly it is you who is misunderstanding something here.

Your scenario is plausible, it's just not group selection. It's K selection. Just because "group" is mentioned in the scenario doesn't make it group selection.

Your definition also makes every adaptation that rescues a group from extinction an example of group selection. What's special about reproductive restraint? Antibiotic resistance is group selection, since otherwise that population would go extinct. Indus... (read more)

Oh, and here's some data on fox reproduction. What would you expect X and Y to be before checking the actual data?

-Of the X to X+1 cubs born to an average fox litter, an average of Y to Y+1 survive the first year.

Windy, according to my logic, yes, to a certain degree all adaptations contribute to some sort of group survival, thus negating the importance of drawing the distinction between group and individual/K selection as some sort of fundamental difference in the mechanics of evolution.

That doesn't mean I'm saying 'group selection' is not a valid area of study, it still needs to be resolved how some adaptations which seem detrimental to the individual end up being good for the individual by proxy of the being good for the group. This is not so much a redefinitio... (read more)

Hi Eliezar

Two questions
1) Which equation in the D.S. Wilson paper are referring to with your C/B < Fst equation formula relates to the paper, this look more like Hmailton's equation which Wilson is, indirectly, criticizing and trying to show that this (kin selection) is a sub-set of multi-level selection. Where are the two levels of selection in this equation which is the key basis for modern group selection formulations?

2) Why not use a criticism of group selection based on claims that are currently being made rather than a discredited scenario (Wy... (read more)

Eliezer's report of Wade's study is not completely accurate. From the link, emphasis mine:

...some of the B populations [the ones selected for low population] enjoy a higher cannibalism rate while other B populations have a longer mean developmental time or a lower average fecundity relative to the controls.... Thus, the decline in mean population size in the B treatment... is the result of both group selection and individual selection within populations favoring characters responsible for low population size....

In summary, these results indicate (i) th

This is just a nitpick: according to The Other Wiki democratic constitutions are older than English kings, the most egregious example being The Solonian Constitution after reform by Cleisthenes, which codified Athenian "democracy". Yes, I know, semantics.

Yudkowsky, is this where your Babyeating aliens from Three Worlds Collide come from?

See my post about an experimental validation of group selection among nightshades.

Suppose an organism of an A genotype emits pheromons depressing development of female reproductive organs in receptive organisms of same population (with an aa genotype), and so gains resources for its own reproduction during a given season, and then during the next season it doesn't emit the pheromon (through some environmental regulation/...) and the homozygotes of the population get a chance to reproduce... Does this count as group selection? It would be still the same species, since the A can receive sperm from any genotype, and use of resources can be... (read more)

Reminds me of one of the early AI research projects using some variety of optimization algorithm to try to "learn" the ability to solve a wide variety of problems in a single program.  Genetic algorithm I think, random mutation and cross-pollination of the programs between the best performers, that kind of thing.

After a while, they noticed that one of the lines that had developed, while not the best at any of the test problems, was second-best at all of them.

Yet when they tried to make it the base of all their next generation...  it didn't work..... (read more)