Lost Purposes

It was in ei­ther kinder­garten or first grade that I was first asked to pray, given a transliter­a­tion of a He­brew prayer. I asked what the words meant. I was told that so long as I prayed in He­brew, I didn’t need to know what the words meant, it would work any­way.

That was the be­gin­ning of my break with Ju­daism.

As you read this, some young man or woman is sit­ting at a desk in a uni­ver­sity, earnestly study­ing ma­te­rial they have no in­ten­tion of ever us­ing, and no in­ter­est in know­ing for its own sake. They want a high-pay­ing job, and the high-pay­ing job re­quires a piece of pa­per, and the piece of pa­per re­quires a pre­vi­ous mas­ter’s de­gree, and the mas­ter’s de­gree re­quires a bach­e­lor’s de­gree, and the uni­ver­sity that grants the bach­e­lor’s de­gree re­quires you to take a class in 12th-cen­tury knit­ting pat­terns to grad­u­ate. So they dili­gently study, in­tend­ing to for­get it all the mo­ment the fi­nal exam is ad­ministered, but still se­ri­ously work­ing away, be­cause they want that piece of pa­per.

Maybe you re­al­ized it was all mad­ness, but I bet you did it any­way. You didn’t have a choice, right?

A re­cent study here in the Bay Area showed that 80% of teach­ers in K-5 re­ported spend­ing less than one hour per week on sci­ence, and 16% said they spend no time on sci­ence. Why? I’m given to un­der­stand the prox­i­mate cause is the No Child Left Be­hind Act and similar leg­is­la­tion. Vir­tu­ally all class­room time is now spent on prepar­ing for tests man­dated at the state or fed­eral level. I seem to re­call (though I can’t find the source) that just tak­ing manda­tory tests was 40% of class­room time in one school.

The old Soviet bu­reau­cracy was fa­mous for be­ing more in­ter­ested in ap­pear­ances than re­al­ity. One shoe fac­tory over­fulfilled its quota by pro­duc­ing lots of tiny shoes. Another shoe fac­tory re­ported cut but unassem­bled leather as a “shoe”. The su­pe­rior bu­reau­crats weren’t in­ter­ested in look­ing too hard, be­cause they also wanted to re­port quota over­fulfill­ments. All this was a great help to the com­rades freez­ing their feet off.

It is now be­ing sug­gested in sev­eral sources that an ac­tual ma­jor­ity of pub­lished find­ings in medicine, though “statis­ti­cally sig­nifi­cant with p<0.05”, are un­true. But so long as p<0.05 re­mains the thresh­old for pub­li­ca­tion, why should any­one hold them­selves to higher stan­dards, when that re­quires big­ger re­search grants for larger ex­per­i­men­tal groups, and de­creases the like­li­hood of get­ting a pub­li­ca­tion? Every­one knows that the whole point of sci­ence is to pub­lish lots of pa­pers, just as the whole point of a uni­ver­sity is to print cer­tain pieces of parch­ment, and the whole point of a school is to pass the manda­tory tests that guaran­tee the an­nual bud­get. You don’t get to set the rules of the game, and if you try to play by differ­ent rules, you’ll just lose.

(Though for some rea­son, physics jour­nals re­quire a thresh­old of p<0.0001. It’s as if they con­ceive of some other pur­pose to their ex­is­tence than pub­lish­ing physics pa­pers.)

There’s choco­late at the su­per­mar­ket, and you can get to the su­per­mar­ket by driv­ing, and driv­ing re­quires that you be in the car, which means open­ing your car door, which needs keys. If you find there’s no choco­late at the su­per­mar­ket, you won’t stand around open­ing and slam­ming your car door be­cause the car door still needs open­ing. I rarely no­tice peo­ple los­ing track of plans they de­vised them­selves.

It’s an­other mat­ter when in­cen­tives must flow through large or­ga­ni­za­tions—or worse, many differ­ent or­ga­ni­za­tions and in­ter­est groups, some of them gov­ern­men­tal. Then you see be­hav­iors that would mark literal in­san­ity, if they were born from a sin­gle mind. Some­one gets paid ev­ery time they open a car door, be­cause that’s what’s mea­surable; and this per­son doesn’t care whether the driver ever gets paid for ar­riv­ing at the su­per­mar­ket, let alone whether the buyer pur­chases the choco­late, or whether the eater is happy or starv­ing.

From a Bayesian per­spec­tive, sub­goals are epiphe­nom­ena of con­di­tional prob­a­bil­ity func­tions. There is no ex­pected util­ity with­out util­ity. How silly would it be to think that in­stru­men­tal value could take on a math­e­mat­i­cal life of its own, leav­ing ter­mi­nal value in the dust? It’s not sane by de­ci­sion-the­o­ret­i­cal crite­ria of san­ity.

But con­sider the No Child Left Be­hind Act. The poli­ti­ci­ans want to look like they’re do­ing some­thing about ed­u­ca­tional difficul­ties; the poli­ti­ci­ans have to look busy to vot­ers this year, not fif­teen years later when the kids are look­ing for jobs. The poli­ti­ci­ans are not the con­sumers of ed­u­ca­tion. The bu­reau­crats have to show progress, which means that they’re only in­ter­ested in progress that can be mea­sured this year. They aren’t the ones who’ll end up ig­no­rant of sci­ence. The pub­lish­ers who com­mis­sion text­books, and the com­mit­tees that pur­chase text­books, don’t sit in the class­rooms bored out of their skulls.

The ac­tual con­sumers of knowl­edge are the chil­dren—who can’t pay, can’t vote, can’t sit on the com­mit­tees. Their par­ents care for them, but don’t sit in the classes them­selves; they can only hold poli­ti­ci­ans re­spon­si­ble ac­cord­ing to sur­face images of “tough on ed­u­ca­tion”. Poli­ti­ci­ans are too busy be­ing re-elected to study all the data them­selves; they have to rely on sur­face images of bu­reau­crats be­ing busy and com­mis­sion­ing stud­ies—it may not work to help any chil­dren, but it works to let poli­ti­ci­ans ap­pear car­ing. Bureau­crats don’t ex­pect to use text­books them­selves, so they don’t care if the text­books are hideous to read, so long as the pro­cess by which they are pur­chased looks good on the sur­face. The text­book pub­lish­ers have no mo­tive to pro­duce bad text­books, but they know that the text­book pur­chas­ing com­mit­tee will be com­par­ing text­books based on how many differ­ent sub­jects they cover, and that the fourth-grade pur­chas­ing com­mit­tee isn’t co­or­di­nated with the third-grade pur­chas­ing com­mit­tee, so they cram as many sub­jects into one text­book as pos­si­ble. Teach­ers won’t get through a fourth of the text­book be­fore the end of the year, and then the next year’s teacher will start over. Teach­ers might com­plain, but they aren’t the de­ci­sion-mak­ers, and ul­ti­mately, it’s not their fu­ture on the line, which puts sharp bounds on how much effort they’ll spend on un­paid al­tru­ism...

It’s amaz­ing, when you look at it that way—con­sider all the lost in­for­ma­tion and lost in­cen­tives—that any­thing at all re­mains of the origi­nal pur­pose, gain­ing knowl­edge. Though many ed­u­ca­tional sys­tems seem to be cur­rently in the pro­cess of col­laps­ing into a state not much bet­ter than noth­ing.

Want to see the prob­lem re­ally solved? Make the poli­ti­ci­ans go to school.

A sin­gle hu­man mind can track a prob­a­bil­is­tic ex­pec­ta­tion of util­ity as it flows through the con­di­tional chances of a dozen in­ter­me­di­ate events—in­clud­ing non­lo­cal de­pen­den­cies, places where the ex­pected util­ity of open­ing the car door de­pends on whether there’s choco­late in the su­per­mar­ket. But or­ga­ni­za­tions can only re­ward to­day what is mea­surable to­day, what can be writ­ten into le­gal con­tract to­day, and this means mea­sur­ing in­ter­me­di­ate events rather than their dis­tant con­se­quences. Th­ese in­ter­me­di­ate mea­sures, in turn, are leaky gen­er­al­iza­tions—of­ten very leaky. Bureau­crats are un­trust­wor­thy ge­nies, for they do not share the val­ues of the wisher.

Miyamoto Musashi said:

“The pri­mary thing when you take a sword in your hands is your in­ten­tion to cut the en­emy, what­ever the means. When­ever you parry, hit, spring, strike or touch the en­emy’s cut­ting sword, you must cut the en­emy in the same move­ment. It is es­sen­tial to at­tain this. If you think only of hit­ting, spring­ing, strik­ing or touch­ing the en­emy, you will not be able ac­tu­ally to cut him. More than any­thing, you must be think­ing of car­ry­ing your move­ment through to cut­ting him. You must thor­oughly re­search this.”

(I wish I lived in an era where I could just tell my read­ers they have to thor­oughly re­search some­thing, with­out giv­ing in­sult.)

Why would any in­di­vi­d­ual lose track of their pur­poses in a sword­fight? If some­one else had taught them to fight, if they had not gen­er­ated the en­tire art from within them­selves, they might not un­der­stand the rea­son for par­ry­ing at one mo­ment, or spring­ing at an­other mo­ment; they might not re­al­ize when the rules had ex­cep­tions, fail to see the times when the usual method won’t cut through.

The es­sen­tial thing in the art of epistemic ra­tio­nal­ity is to un­der­stand how ev­ery rule is cut­ting through to the truth in the same move­ment. The cor­re­spond­ing es­sen­tial of prag­matic ra­tio­nal­ity—de­ci­sion the­ory, ver­sus prob­a­bil­ity the­ory—is to always see how ev­ery ex­pected util­ity cuts through to util­ity. You must thor­oughly re­search this.

C. J. Cher­ryh said:

“Your sword has no blade. It has only your in­ten­tion. When that goes astray you have no weapon.”

I have seen many peo­ple go astray when they wish to the ge­nie of an imag­ined AI, dream­ing up wish af­ter wish that seems good to them, some­times with many patches and some­times with­out even that pre­tense of cau­tion. And they don’t jump to the meta-level. They don’t in­stinc­tively look-to-pur­pose, the in­stinct that started me down the track to athe­ism at the age of five. They do not ask, as I re­flex­ively ask, “Why do I think this wish is a good idea? Will the ge­nie judge like­wise?” They don’t see the source of their judg­ment, hov­er­ing be­hind the judg­ment as its gen­er­a­tor. They lose track of the ball; they know the ball bounced, but they don’t in­stinc­tively look back to see where it bounced from—the crite­rion that gen­er­ated their judg­ments.

Like­wise with peo­ple not au­to­mat­i­cally notic­ing when sup­pos­edly self­ish peo­ple give al­tru­is­tic ar­gu­ments in fa­vor of self­ish­ness, or when sup­pos­edly al­tru­is­tic peo­ple give self­ish ar­gu­ments in fa­vor of al­tru­ism.

Peo­ple can han­dle goal-track­ing for driv­ing to the su­per­mar­ket just fine, when it’s all in­side their own heads, and no ge­nies or bu­reau­cra­cies or philoso­phies are in­volved. The trou­ble is that real civ­i­liza­tion is im­mensely more com­pli­cated than this. Dozens of or­ga­ni­za­tions, and dozens of years, in­ter­vene be­tween the child suffer­ing in the class­room, and the new-minted col­lege grad­u­ate not be­ing very good at their job. (But will the in­ter­viewer or man­ager no­tice, if the col­lege grad­u­ate is good at look­ing busy?) With ev­ery new link that in­ter­venes be­tween the ac­tion and its con­se­quence, in­ten­tion has one more chance to go astray. With ev­ery in­ter­ven­ing link, in­for­ma­tion is lost, in­cen­tive is lost. And this both­ers most peo­ple a lot less than it both­ers me, or why were all my class­mates will­ing to say prayers with­out know­ing what they meant? They didn’t feel the same in­stinct to look-to-the-gen­er­a­tor.

Can peo­ple learn to keep their eye on the ball? To keep their in­ten­tion from go­ing astray? To never spring or strike or touch, with­out know­ing the higher goal they will com­plete in the same move­ment? Peo­ple do of­ten want to do their jobs, all else be­ing equal. Can there be such a thing as a sane cor­po­ra­tion? A sane civ­i­liza­tion, even? That’s only a dis­tant dream, but it’s what I’ve been get­ting at with all these blog posts on the flow of in­ten­tions (aka ex­pected util­ity, aka in­stru­men­tal value) with­out los­ing pur­pose (aka util­ity, aka ter­mi­nal value). Can peo­ple learn to feel the flow of par­ent goals and child goals? To know con­sciously, as well as im­plic­itly, the dis­tinc­tion be­tween ex­pected util­ity and util­ity?

Do you care about threats to your civ­i­liza­tion? The worst metathreat to com­plex civ­i­liza­tion is its own com­plex­ity, for that com­pli­ca­tion leads to the loss of many pur­poses.

I look back, and I see that more than any­thing, my life has been driven by an ex­cep­tion­ally strong ab­hor­rence to lost pur­poses. I hope it can be trans­formed to a learn­able skill.