Religion as Goodhart

This post is in­spired by the SSC Book Re­view: The Se­cret of Our Suc­cess. To (dou­ble- or triple-)quote the rele­vant parts:

the best hunt­ing strat­egy re­quires randomizing


If you’re gen­er­ally pre­dictable – and Scott Aaron­son says you are – then out­sourc­ing your de­ci­sion to weird birds might be the best way to go.

This is a some­what gen­eral pat­tern: to deal with smart ad­ver­saries effec­tively, ran­dom­iza­tion is, on av­er­age, the best strat­egy. Match­ing pen­nies and Rock/​pa­per/​scis­sors are some clas­sic ex­am­ples. Also, hu­mans (and prob­a­bly an­i­mals in gen­eral) are ap­par­ently poor ran­dom­iz­ers. I as­sume be­cause a built-in biolog­i­cal ran­dom num­ber gen­er­a­tor of ad­e­quate qual­ity is ex­pen­sive. I don’t know if this is a good as­sump­tion. A quick on­line search does not re­fute or con­firm this, and ran­dom­ness is ev­ery­where at the cel­lu­lar level, but maybe not at the level of the sin­gle or­ganisms.

For­tu­nately, true ran­dom­ness is ev­ery­where out­side in the phys­i­cal pro­cesses, and is cheap to ac­cess, and the au­gury and div­ina­tion men­tioned in the re­viewed book are good ex­am­ples of that. So, rely­ing on ex­ter­nal ran­dom­ness might be an evolu­tion­ary adap­ta­tion that im­proves sur­vival. (Is this so? Then we would see similar be­hav­ior in other an­i­mals’ strate­gies for deal­ing with ad­ver­sity, be it hunt­ing, gath­er­ing or some­thing else. Do we?)

Of course, the next step, con­sciously un­der­stand­ing what ran­dom­ness is and why it is use­ful, is bor­der­line im­pos­si­ble for any an­i­mal but a rather ad­vanced hu­man, and even then it’s iffy: even statis­ti­ci­ans are not im­mune to be­hav­ioral rit­u­als. And given our awe­some pow­ers of ra­tio­nal­iza­tion, we jus­tify the ran­dom­iza­tion rit­u­als as pre­dic­tive, not just use­ful.

And so we have a Good­hart-like situ­a­tion, where a proxy be­comes the goal, and is no longer a use­ful proxy:

  • The “true” goal is to im­prove the odds of find­ing or cul­ti­vat­ing a re­source nec­es­sary for sur­vival.

  • The cho­sen proxy is to rely on an ex­ter­nal ran­dom­izer.

  • The in­ter­nal jus­tifi­ca­tion for it is that it has larger pre­dic­tive pow­ers then the per­son’s logic or in­tu­ition.

  • Thus be­liev­ing in an ex­ter­nal power that is larger than one­self has evolu­tion­ary ad­van­tages and can stick, so the proxy be­comes the goal, even though it gives lit­tle or no ad­van­tage in most situ­a­tions that do not re­quire ran­dom­iza­tion.

  • Ergo, “a God-shaped hole” in one’s heart.

The above is largely a spec­u­la­tion and is miss­ing a fair bit in or­der to be­come a qual­ity model of how the need to be­lieve in the su­per­nat­u­ral may have been se­lected as an evolu­tion­ary ad­van­tage. But I don’t see any­thing im­me­di­ately wrong with it. It also im­me­di­ately makes a num­ber of pre­dic­tions, such as some non-hu­man an­i­mals us­ing ex­ter­nal ran­dom events to guide their be­hav­ior, in­stead of bas­ing it on more de­ter­minis­tic fac­tors, or on an in­ter­nal gen­er­a­tor of ran­dom­ness.

It is worth not­ing that this might not truly be a Good­hart-like situ­a­tion, since the proxy is still use­ful in its origi­nal con­text, and, po­ten­tially, in other con­texts, such as com­mu­nity build­ing.

I have not put any re­search into whether this has been dis­cussed be­fore, here or el­se­where, my apolo­gies if this is a rep­e­ti­tion of some­thing already well known.