[Question] How can guesstimates work?

[Epistemic sta­tus: back­ground is very hand-wavy, but I’d rather post a rough ques­tion than no ques­tion at all. I’m very con­fi­dent that the two in­gre­di­ents—illeg­ible cul­tural evolu­tion and guessti­ma­tion—are real and im­por­tant things. Though the re­la­tion be­tween the two is more un­cer­tain. I’m not that sur­prised if my ques­tion ends up con­fused and dis­solved rather than solved by an­swers.]

For a large part of hu­man his­tory, our lives were dic­tated by cul­tural norms and pro­cesses which ap­peared ar­bi­trary, yet could have fatal con­squences if de­parted from. (C.f. SSC on The Se­cret of Our Suc­cess, which will be as­sumed back­ground knowl­edge for this ques­tion.)

To­day, we live in a world where you can achieve huge gains if you sim­ply “shut up and mul­ti­ply”. The world seems leg­ible—I can roughly pre­dict how many planes fly ev­ery day by mul­ti­ply­ing a hand­ful rough num­bers. And the world seems ed­itable—peo­ple who like to cook of­ten im­pro­vise: ex­chang­ing, adding and re­mov­ing in­gre­di­ents. And this seems fine. It cer­tainly doesn’t kill them. Hugely suc­ces­ful com­pa­nies are built around the prin­ci­ple of “just try things un­til some­thing breaks and then try again and im­prove”.

I still think there are still large amounts of illeg­ible cul­tural knowl­edge en­coded in in­sti­tu­tions, tra­di­tions, norms, etc. But some­thing still seems very differ­ent from the hor­ror sto­ries of epistemic learned hel­pless­ness Scott shared.

What changed about the world to make this pos­si­ble? How can guessti­mates work?

Some hy­pothe­ses (none of which I’d put more than 15% on atm):

  • Al­most all im­por­tant as­pects of our lives our gov­erned by some kind of tech­nol­ogy that we built (ta­bles, air­planes, com­put­ers, rugs, restau­rants, microwaves, le­gal con­tracts, clocks, beds, clothes, … and so on and so forth). Tech­nolog­i­cal de­vel­op­ment out­paced cul­tural evolu­tion. The mod­ern world is more leg­ible and ed­itable for the same rea­son that a code­base is more leg­ible and ed­itable than DNA.

  • Most sys­tems that gov­ern our lives have been op­ti­mised way harder since the in­dus­trial rev­olu­tion and the abil­ity to achieve economies-of-scale in a global mar­ket. Things are gen­er­ally closer to op­ti­mal equil­ibria, and equil­ibria are more leg­ible/​pre­dictable than non-quil­ibria.

  • The things we re­ally care about to­day of­ten fol­low very heavy-tailed dis­tri­bu­tions. Hence even a very rough guessti­mate is like to get the or­der­ing of choices cor­rect. (This can’t be the whole story, be­cause guessti­mates also seem to work in pretty Gaus­sian do­mains.)

  • We just have bet­ter med­i­cal and welfare sys­tems which al­low peo­ple to take more risks.