The Tragedy of Group Selectionism

Be­fore 1966, it was not un­usual to see se­ri­ous biol­o­gists ad­vo­cat­ing evolu­tion­ary hy­pothe­ses that we would now re­gard as mag­i­cal think­ing. Th­ese mud­dled no­tions played an im­por­tant his­tor­i­cal role in the de­vel­op­ment of later evolu­tion­ary the­ory, er­ror call­ing forth cor­rec­tion; like the folly of English kings pro­vok­ing into ex­is­tence the Magna Carta and con­sti­tu­tional democ­racy.

As an ex­am­ple of ro­mance, Vero Wynne-Ed­wards, Warder Allee, and J. L. Br­ere­ton, among oth­ers, be­lieved that preda­tors would vol­un­tar­ily re­strain their breed­ing to avoid over­pop­u­lat­ing their habitat and ex­haust­ing the prey pop­u­la­tion.

But evolu­tion does not open the floodgates to ar­bi­trary pur­poses. You can­not ex­plain a rat­tlesnake’s rat­tle by say­ing that it ex­ists to benefit other an­i­mals who would oth­er­wise be bit­ten. No out­side Evolu­tion Fairy de­cides when a gene ought to be pro­moted; the gene’s effect must some­how di­rectly cause the gene to be more preva­lent in the next gen­er­a­tion. It’s clear why our hu­man sense of aes­thet­ics, wit­ness­ing a pop­u­la­tion crash of foxes who’ve eaten all the rab­bits, cries “Some­thing should’ve been done!” But how would a gene com­plex for re­strain­ing re­pro­duc­tion—of all things!—cause it­self to be­come more fre­quent in the next gen­er­a­tion?

A hu­man be­ing de­sign­ing a neat lit­tle toy ecol­ogy—for en­ter­tain­ment pur­poses, like a model railroad—might be an­noyed if their painstak­ingly con­structed fox and rab­bit pop­u­la­tions self-de­struc­ted by the foxes eat­ing all the rab­bits and then dy­ing of star­va­tion them­selves. So the hu­man would tin­ker with the toy ecol­ogy—a fox-breed­ing-re­strainer is the ob­vi­ous solu­tion that leaps to our hu­man minds—un­til the ecol­ogy looked nice and neat. Na­ture has no hu­man, of course, but that needn’t stop us—now that we know what we want on aes­thetic grounds, we just have to come up with a plau­si­ble ar­gu­ment that per­suades Na­ture to want the same thing on evolu­tion­ary grounds.

Ob­vi­ously, se­lec­tion on the level of the in­di­vi­d­ual won’t pro­duce in­di­vi­d­ual re­straint in breed­ing. In­di­vi­d­u­als who re­pro­duce un­re­strainedly will, nat­u­rally, pro­duce more offspring than in­di­vi­d­u­als who re­strain them­selves.

(Ad­den­dum: In­di­vi­d­ual se­lec­tion will not pro­duce in­di­vi­d­ual sac­ri­fice of breed­ing op­por­tu­ni­ties. In­di­vi­d­ual se­lec­tion can cer­tainly pro­duce in­di­vi­d­u­als who, af­ter ac­quiring all available re­sources, use those re­sources to pro­duce 4 big eggs in­stead of 8 small eggs—not to con­serve so­cial re­sources, but be­cause that is the in­di­vi­d­ual sweet spot for num­ber of eggs * egg sur­vival prob­a­bil­ity. This does not get rid of the com­mons prob­lem.)

But sup­pose that the species pop­u­la­tion was bro­ken up into sub­pop­u­la­tions, which were mostly iso­lated, and only oc­ca­sion­ally in­ter­bred. Then, surely, sub­pop­u­la­tions that re­strained their breed­ing would be less likely to go ex­tinct, and would send out more mes­sen­gers, and cre­ate new colonies to rein­habit the ter­ri­to­ries of crashed pop­u­la­tions.

The prob­lem with this sce­nario wasn’t that it was math­e­mat­i­cally im­pos­si­ble. The prob­lem was that it was pos­si­ble but very difficult.

The fun­da­men­tal prob­lem is that it’s not only re­strained breed­ers who reap the benefits of re­strained breed­ing. If some foxes re­frain from spawn­ing cubs who eat rab­bits, then the un­eaten rab­bits don’t go to only cubs who carry the re­strained-breed­ing adap­ta­tion. The un­re­strained foxes, and their many more cubs, will hap­pily eat any rab­bits left un­hunted. The only way the re­strain­ing gene can sur­vive against this pres­sure, is if the benefits of re­straint prefer­en­tially go to re­strain­ers.

Speci­fi­cally, the re­quire­ment is C/​B < FST where C is the cost of al­tru­ism to the donor, B is the benefit of al­tru­ism to the re­cip­i­ent, and FST is the spa­tial struc­ture of the pop­u­la­tion: the av­er­age re­lat­ed­ness be­tween a ran­domly se­lected or­ganism and its ran­domly se­lected neigh­bor, where a “neigh­bor” is any other fox who benefits from an al­tru­is­tic fox’s re­straint. (I be­lieve this is a deriva­tion with differ­ent sym­bols, best one I could find on­line.)

So is the cost of re­strained breed­ing suffi­ciently small, and the em­piri­cal benefit of less famine suffi­ciently large, com­pared to the em­piri­cal spa­tial struc­ture of fox pop­u­la­tions and rab­bit pop­u­la­tions, that the group se­lec­tion ar­gu­ment can work?

The math sug­gests this is pretty un­likely. In this simu­la­tion, for ex­am­ple, the cost to al­tru­ists is 3% of fit­ness, pure al­tru­ist groups have a fit­ness twice as great as pure self­ish groups, the sub­pop­u­la­tion size is 25, and 20% of all deaths are re­placed with mes­sen­gers from an­other group: the re­sult is poly­mor­phic for self­ish­ness and al­tru­ism. If the sub­pop­u­la­tion size is dou­bled to 50, self­ish­ness is fixed; if the cost to al­tru­ists is in­creased to 6%, self­ish­ness is fixed; if the al­tru­is­tic benefit is de­creased by half, self­ish­ness is fixed or in large ma­jor­ity. Neigh­bor­hood-groups must be very small, with only around 5 mem­bers, for group se­lec­tion to op­er­ate when the cost of al­tru­ism ex­ceeds 10%. This doesn’t seem plau­si­bly true of foxes re­strain­ing their breed­ing.

You can guess by now, I think, that the group se­lec­tion­ists ul­ti­mately lost the sci­en­tific ar­gu­ment. The kicker was not the math­e­mat­i­cal ar­gu­ment, but em­piri­cal ob­ser­va­tion: foxes didn’t re­strain their breed­ing (I for­get the ex­act species of dis­pute; it wasn’t foxes and rab­bits), and in­deed, preda­tor-prey sys­tems crash all the time. 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. So why is this now-his­tor­i­cal sci­en­tific dis­pute wor­thy ma­te­rial for Over­com­ing Bias?

A decade af­ter the con­tro­versy, a biol­o­gist had a fas­ci­nat­ing idea. The math­e­mat­i­cal con­di­tions for group se­lec­tion over­com­ing in­di­vi­d­ual se­lec­tion were too ex­treme to be found in Na­ture. Why not cre­ate them ar­tifi­cially, in the lab­o­ra­tory? Michael J. Wade pro­ceeded to do just that, re­peat­edly se­lect­ing pop­u­la­tions of in­sects for low num­bers of adults per sub­pop­u­la­tion. And what was the re­sult? Did the in­sects re­strain their breed­ing and live in quiet peace with enough food for all?

No; the adults adapted to can­ni­bal­ize eggs and lar­vae, es­pe­cially fe­male lar­vae.

Of course se­lect­ing for small sub­pop­u­la­tion sizes would not se­lect for in­di­vi­d­u­als who re­strained their own breed­ing; it would se­lect for in­di­vi­d­u­als who ate other in­di­vi­d­u­als’ chil­dren. Espe­cially the girls.

Once you have that ex­per­i­men­tal re­sult in hand—and it’s mas­sively ob­vi­ous in ret­ro­spect—then it sud­denly be­comes clear how the origi­nal group se­lec­tion­ists al­lowed ro­man­ti­cism, a hu­man sense of aes­thet­ics, to cloud their pre­dic­tions of Na­ture.

This is an archety­pal ex­am­ple of a missed Third Alter­na­tive, re­sult­ing from a ra­tio­nal­iza­tion of a pre­de­ter­mined bot­tom line which pro­duced a fake jus­tifi­ca­tion and then mo­ti­vat­edly stopped. The group se­lec­tion­ists didn’t start with clear, fresh minds, hap­pen upon the idea of group se­lec­tion, and neu­trally ex­trap­o­late for­ward the prob­a­ble out­come. They started out with the beau­tiful idea of fox pop­u­la­tions vol­un­tar­ily re­strain­ing their re­pro­duc­tion to what the rab­bit pop­u­la­tion would bear, Na­ture in perfect har­mony; then they searched for a rea­son why this would hap­pen, and came up with the idea of group se­lec­tion; then, since they knew what they wanted the out­come of group se­lec­tion to be, they didn’t look for any less beau­tiful and aes­thetic adap­ta­tions that group se­lec­tion would be more likely to pro­mote in­stead. If they’d re­ally been try­ing to calmly and neu­trally pre­dict the re­sult of se­lect­ing for small sub­pop­u­la­tion sizes re­sis­tant to famine, they would have thought of can­ni­bal­iz­ing other or­ganisms’ chil­dren or some similarly “ugly” out­come—long be­fore they imag­ined any­thing so evolu­tion­ar­ily outré as in­di­vi­d­ual re­straint in breed­ing!

This also illus­trates the point I was try­ing to make in Ein­stein’s Ar­ro­gance: With large an­swer spaces, nearly all of the real work goes into pro­mot­ing one pos­si­ble an­swer to the point of be­ing sin­gled out for at­ten­tion. If a hy­poth­e­sis is im­prop­erly pro­moted to your at­ten­tion—your sense of aes­thet­ics sug­gests a beau­tiful way for Na­ture to be, and yet nat­u­ral se­lec­tion doesn’t in­volve an Evolu­tion Fairy who shares your ap­pre­ci­a­tion—then this alone may seal your doom, un­less you can man­age to clear your mind en­tirely and start over.

In prin­ci­ple, the world’s stupi­dest per­son may say the Sun is shin­ing, but that doesn’t make it dark out. Even if an an­swer is sug­gested by a lu­natic on LSD, you should be able to neu­trally calcu­late the ev­i­dence for and against, and if nec­es­sary, un-be­lieve.

In prac­tice, the group se­lec­tion­ists were doomed be­cause their bot­tom line was origi­nally sug­gested by their sense of aes­thet­ics, and Na­ture’s bot­tom line was pro­duced by nat­u­ral se­lec­tion. Th­ese two pro­cesses had no prin­ci­pled rea­son for their out­puts to cor­re­late, and in­deed they didn’t. All the fu­ri­ous ar­gu­ment af­ter­ward didn’t change that.

If you start with your own de­sires for what Na­ture should do, con­sider Na­ture’s own ob­served rea­sons for do­ing things, and then ra­tio­nal­ize an ex­tremely per­sua­sive ar­gu­ment for why Na­ture should pro­duce your preferred out­come for Na­ture’s own rea­sons, then Na­ture, alas, still won’t listen. The uni­verse has no mind and is not sub­ject to clever poli­ti­cal per­sua­sion. You can ar­gue all day why grav­ity should re­ally make wa­ter flow up­hill, and the wa­ter just ends up in the same place re­gard­less. It’s like the uni­verse plain isn’t listen­ing. J. R. Mol­loy said: “Na­ture is the ul­ti­mate bi­got, be­cause it is ob­sti­nately and in­tol­er­antly de­voted to its own prej­u­dices and ab­solutely re­fuses to yield to the most per­sua­sive ra­tio­nal­iza­tions of hu­mans.”

I of­ten recom­mend evolu­tion­ary biol­ogy to friends just be­cause the mod­ern field tries to train its stu­dents against ra­tio­nal­iza­tion, er­ror call­ing forth cor­rec­tion. Physi­cists and elec­tri­cal en­g­ineers don’t have to be care­fully trained to avoid an­thro­po­mor­phiz­ing elec­trons, be­cause elec­trons don’t ex­hibit mindish be­hav­iors. Nat­u­ral se­lec­tion cre­ates pur­pose­ful­nesses which are alien to hu­mans, and stu­dents of evolu­tion­ary the­ory are warned ac­cord­ingly. It’s good train­ing for any thinker, but it is es­pe­cially im­por­tant if you want to think clearly about other weird mindish pro­cesses that do not work like you do.