Constructing fictional eugenics (LW edition)

Yvain asked:

So if you had to de­sign a eu­gen­ics pro­gram, how would you do it? Be cre­ative.

I’m ask­ing be­cause I’m work­ing on writ­ing about a fic­tional so­ciety that prac­tices eu­gen­ics. I want them to be in­ter­est­ing and sym­pa­thetic, and not to im­me­di­ately pat­tern-match to a dystopia that kills ev­ery­one who doesn’t look ex­actly al­ike.

My re­ply was too long for LiveJour­nal, so I’m post­ing it here:

1. The real step 1 in any pro­gram like this would be to buy the 3 best mod­ern text­books on an­i­mal breed­ing and read them. (My grand­father is a re­searcher in this field so I’m un­usu­ally aware that it ex­ists.)

2. If you give me ge­netic se­lec­tion on mul­ti­ple pos­si­ble em­bryos where I can read off the genome of each one, I can do much bet­ter, much faster, than if I’m only al­lowed to look at the mother and father’s genome and pre­dict on that ba­sis. If I can only look at the mother and father’s rel­a­tives and life achieve­ments, I do worse, but mod­ern tech is very rapidly ad­vanc­ing to be able to read off the par­ents’ genome cheaply.

3. If so­ciety’s util­ity has a large com­po­nent for ge­nius pro­duc­tion, then you prob­a­bly want a very di­verse mix of differ­ent high-IQ genes com­bined into differ­ent geno­types and phe­no­types. (Although some re­cent re­search sug­gests that the most im­por­tant thing for IQ may be avoid­ing mu­ta­tional load, i.e., the modal genome would be su­per-von-Neu­mann. Even so, we’d want a di­verse mix of ev­ery­thing else cog­ni­tive that wasn’t about modal­ity.)

4. Do­ing a Bayesian value-of-in­for­ma­tion calcu­la­tion on rare alle­les and po­ten­tially in­ter­est­ing allele com­bi­na­tions will au­to­mat­i­cally in­clude a value for di­ver­sity into your eu­genic pro­gram, based on the value of pro­mot­ing a gene /​ combo in much larger num­bers if that gene or gene combo is found to be suc­cess­ful. You would get much *more* in­ter­est­ing di­ver­sity in the next gen­er­a­tion au­to­mat­i­cally, as many pre­vi­ously low-fre­quency alle­les were com­bined in greater num­bers and greater di­ver­sity than be­fore. *Not* do­ing a value-of-info calcu­la­tion ac­counts for a lot of the dystopic load of alleged dystopias.

5. The ob­vi­ous ba­sic in­stru­ment in a so­ciety de­picted as well-in­ten­tioned would be an eco­nomic policy of try­ing to in­ter­nal­ize the ex­ter­nal­ities of a child, just like a well-in­ten­tioned so­ciety might try to in­ter­nal­ize the ex­ter­nal­ities of e.g. car­bon diox­ide emis­sions, in­stead of reg­u­lat­ing/​cap­ping them di­rectly, in or­der to max­i­mize net so­cial welfare. There would be a tax or benefit based on how much your child is ex­pected to cost so­ciety (not just gov­ern­men­tal costs in the form of health care, school­ing etc., but costs to so­ciety in gen­eral, in­clud­ing fore­gone la­bor of a work­ing par­ent, etc.) and how much that child is ex­pected to benefit so­ciety (not life­time tax rev­enue or life­time earn­ings, but life­time value gen­er­ated—most eco­nomic ac­tors only cap­ture a frac­tion of the value they cre­ate). If it looks like you’re go­ing to have a valuable child, you get your benefit in the form of a large cash bonus up-front (love that hy­per­bolic dis­count­ing) and lots of free child­care so you can go on hav­ing more chil­dren. The mar­keted so­cial goal would be to avert the mod­ern trope where par­ent­hood is this dread­ful bur­den­some in­con­ve­nience com­pared to play­ing video games, and this is bad for so­ciety be­cause so­ciety runs out of valuable fu­ture work­ers whose benefits-to-so­ciety the par­ents mostly don’t cap­ture. Prob­a­bly the hard part from a mar­ket­ing stand­point would be the pro­posal to do ac­tual ge­netic calcu­la­tions, even if it’s to allegedly in­crease so­cial benefit and pre­vent the sys­tem from be­ing “ex­ploited” (i.e. go­ing dys­genic-Malthu­sian).

6. As sug­gested in an ear­lier com­ment, fi­nan­cial­iz­ing pro­gres­sive shares of fu­ture in­come (as di­verted from tax streams, maybe) is an ob­vi­ous way to pri­va­tize pre­dic­tion, but only of tax streams, or at best rev­enue earned by the prospec­tive in­di­vi­d­ual. (I hadn’t thought of this un­til I read that com­ment, so credit where it’s due.)

7. Taxes on ex­pected-nega­tive kids are more icky but would still have the ob­vi­ous eco­nomic jus­tifi­ca­tion. A nicer-sound­ing way of fram­ing it would be re­quiring par­ents to post bond cor­re­spond­ing to the baseline gov­ern­ment cost of each child in school­ing and health­care, with ex­pected value po­ten­tially helping to make up the bond. An in­ter­est­ing ques­tion is whether any­one would re­ally work out to ex­pected-net-nega­tive un­der this sys­tem, which ques­tion is iso­mor­phic to ask­ing whether it ever makes self­ish sense for a coun­try to re­strict im­mi­gra­tion. But adding at least some bur­den here makes sense from a cog­ni­tive per­spec­tive, be­cause adding a cost is bet­ter at shap­ing be­hav­ior than adding a po­ten­tially fore­gone benefit.

8. The in­cen­tive for e.g. tak­ing ad­van­tage of sperm banks is au­to­matic in this sys­tem—you can ei­ther pay a bunch of money to have a kid with your cur­rent hus­band, or you can be paid thou­sands of dol­lars and get free child care to be in­sem­i­nated by the sperm of a No­bel win­ner who never had to diet. I think that, in prac­tice, the ba­sic test of a sys­tem like this would be whether it could get peo­ple to go over the in­con­ve­nience thresh­old of ac­tu­ally us­ing sperm banks and egg donors.

9. More in­ter­est­ingly, there’s a built-in in­cen­tive for most peo­ple to have daugh­ters rather than sons un­der this sys­tem. If we take the ex­pected ex­ter­nal­ities of grand­chil­dren into ac­count in calcu­lat­ing the ex­pected ex­ter­nal­ities of a child, then daugh­ters can bear chil­dren us­ing the best sperm via gene banks, while men have a harder time get­ting at the best eggs, mak­ing the grand­chil­dren of daugh­ters much more valuable if you’ll as­sume they’ll all be No­bel-lau­re­ate-de­scen­dants. Daugh­ters also add more marginal chil­dren to so­ciety than sons, since adding an­other son does not in­crease the marginal re­pro­duc­tive ca­pac­ity of so­ciety un­less sin­gle women aren’t will­ing to re­pro­duce us­ing sperm banks (even tak­ing into ac­count sub­si­dized child­care) and the polyamory fac­tor has gone over what women with chil­dren are will­ing to tol­er­ate. So if grand­chil­dren are net pos­i­tive, daugh­ters are more marginally valuable to so­ciety un­til the sex ra­tio has gone well over 1:1. This is leav­ing aside gen­er­ally larger crim­i­nal down­sides of men, the fact that men do worse in school (which may be a mere ar­ti­fact of our hor­ror of a school sys­tem), and so on. How­ever, if the sex ra­tio be­comes very ex­treme and the sys­tem is sup­posed to stick around for many gen­er­a­tions, then most of the males gen­er­ated will be by peo­ple defy­ing sys­tem in­cen­tives; and un­less very few women re­pro­duce with those males, there will be a large se­lec­tive ad­van­tage for hav­ing sons out­side the sys­tem. I.e., the sys­tem will be se­lect­ing for those who defy its in­cen­tives, which is a key de­sign crite­rion for avoid­ing. (Though on yet fur­ther re­flec­tion, if there are many males with sub­op­ti­mal ge­net­ics be­ing pro­duced and then re­pro­duc­ing, child-value calcu­la­tions would rapidly yield the so­cial ad­vice to start birthing more above-av­er­age males even if they won’t win the sperm-bank con­test; and if women have a strong prefer­ence for pre­sent fathers, you could di­rectly calcu­late that as so­cial value as well as a fac­tor in calcu­lat­ing ex­pected ge­netic im­pact of males.)

10. In the end, all of this just adds up to, “If you can cor­rectly in­ter­nal­ize these ex­ter­nal­ities, the fol­low­ing so­cial welfare fac­tor will be in­creased...” and the key part is of course that “If”.

(See Yvain’s post and com­ments as well.)