Moral Mazes and Short Termism

Link post

Pre­vi­ously: Short Ter­mism and Quotes from Mo­ral Mazes

Epistemic Sta­tus: Long term

My list of quotes from moral mazes has a sec­tion of twenty de­voted to short term think­ing. It fits with, and gives in­ter­nal gears and color to, my pre­vi­ous un­der­stand­ing of of the prob­lem of short ter­mism.

Much of what we think of as a Short Term vs. Long Term is­sue is ac­tu­ally an ad­ver­sar­ial Good­hart’s Law prob­lem, or a leg­i­bil­ity vs. illeg­i­bil­ity prob­lem, at the ob­ject level, that then be­comes a short vs. long term is­sue at higher lev­els. When a man­ager milks a plant (see quotes 72, 73, 78 and 79) they are not pri­mar­ily trad­ing long term as­sets for short term as­sets. Rather, they are trad­ing un­mea­sured as­sets for mea­sured as­sets (see 67 and 69).

This is why you can have com­pa­nies like Ama­zon, Uber or Tesla get high val­u­a­tions. They hit leg­ible short-term met­rics that rep­re­sent long-term growth. A start-up gets re­warded for their own sort of leg­ible short-term in­di­ca­tors of progress and suc­cess, and of the qual­ity of team and there­fore po­ten­tial for fu­ture suc­cess. Whereas other com­pa­nies, that are not based on growth, re­port huge pres­sure to hit profit num­bers.

The over­whelming ob­ject level pres­sure to­wards leg­ible short-term suc­cess, what­ever that means in con­text, comes from be­ing judged in the short term on one’s suc­cess, and hav­ing that judg­ment be­ing more im­por­tant than ob­ject-level long term suc­cess.

The eas­iest way for this to be true is not to care about ob­ject-level long term suc­cess. If you’re gone be­fore the long term, and no one traces the long term back to you, why do you care what hap­pens? That is ex­actly the situ­a­tion the man­agers face in Mo­ral Mazes (see 64, 65, 70, 71, 74 and 83, and for a non-man­ager very clean ex­am­ple see 77). In par­tic­u­lar:

74. We’re judged on the short-term be­cause ev­ery­body changes their jobs so fre­quently.


64. The ideal situ­a­tion, of course, is to end up in a po­si­tion where one can fire one’s suc­ces­sors for one’s own pre­vi­ous mis­takes.

Al­most as good as hav­ing a des­ig­nated scape­goat is to have already sold the com­pany or found em­ploy­ment el­se­where, ren­der­ing your prob­lems some­one else’s prob­lems.

The other way to not care is for the short-term eval­u­a­tion of one’s suc­cess or failure to im­pact long-term suc­cess. If not hit­ting a short-term num­ber gets you fired, or pre­vents your com­pany from get­ting ac­cept­able terms on fi­nanc­ing or gets you bought out, then the long term will get ne­glected. The net pre­sent value pay­off for look­ing good, which can then be rein­vested, makes it look like by far the best long term in­vest­ment around.

Thus we have this prob­lem at ev­ery level of man­age­ment ex­cept the top. But for the top to ac­tu­ally be the top, it needs to not be an­swer­ing to the stock mar­ket or cap­i­tal mar­kets, or oth­er­wise care what oth­ers think – even with­out ex­plicit ver­dicts, this can be as hard to root out as need­ing the per­cep­tion of a bright fu­ture to at­tract and keep qual­ity em­ploy­ees and keep up morale. So we al­most always have it at the top as well. Each level is dis­tort­ing things for the level above, and push­ing these dis­torted pri­ori­ties down to get to the next move in a gi­ant game of ad­ver­sar­ial tele­phone (see sec­tion A of quotes for how hi­er­ar­chy works).

This re­sults in a cor­po­ra­tion that acts in var­i­ous short-term ways, some of which make sense for it, some of which are the re­sult of in­ter­nal con­flicts.

Why isn’t this out-com­peted? Why don’t the cor­po­ra­tions that do less of this drive the ones that do more of it out of the mar­ket?

On the level of cor­po­ra­tions do­ing this di­rect from the top, of­ten these ac­tions are a re­sponse to the in­cen­tives the cor­po­ra­tion faces. In those cases, there is no rea­son to ex­pect such ac­tions to be out-com­peted.

In other cases, the in­cen­tives of the CEO and top man­age­ment are twisted but the cor­po­ra­tion’s in­cen­tives are not. One would cer­tainly ex­pect those cor­po­ra­tions that avoid this to do bet­ter. But these mis­matches are the nat­u­ral con­se­quence of putting some­one in charge who does not per­ma­nently own the com­pany. Thus, dual class share struc­tures be­com­ing pop­u­lar to re­store skin in the cor­rect game. Some of the lower-down is­sues can be made less bad by re­mov­ing the ones at the top, but the prob­lem does not go away, and what sources I have in­side ma­jor tech com­pa­nies in­clud­ing Google match this model.

There is also the ten­dency of these dy­nam­ics to arise over time. Those who play the power game tend to out­perform those who do not play it bar­ring con­stant vigilance and a will­ing­ness to sac­ri­fice. As those play­ers out­perform, they cause other power play­ers to out­perform more, be­cause they pre­fer and fa­vor such other play­ers, and fa­vor rules that fa­vor such play­ers. This is es­pe­cially pow­er­ful for any­one be­low them in the hi­er­ar­chy. An in­fected CEO, who can in­stall their own peo­ple, can quickly be game over on its own, and out­side CEOs are brought in of­ten.

Thus, even if the sys­tem causes the cor­po­ra­tion to un­der­perform, it still spreads, like a meme that in­fects the host, caus­ing the host to pri­ori­tize spread­ing the meme, while re­duc­ing re­pro­duc­tive fit­ness. The big­ger the or­ga­ni­za­tion, the harder it is to re­main un­in­fected. Be­ing able to be tem­porar­ily less bur­dened by such is­sues is one of the big ad­van­tages new en­trants have.

One could even say that yes, they do get wiped out by this, but it’s not that fast, be­cause it takes a while for this to rise to the level of a pri­mary de­ter­min­ing fac­tor in out­comes. And there are big­ger things to worry about. It’s short ter­mism, so that isn’t too sur­pris­ing.

A big pres­sure that causes these in­fec­tions is that busi­ness is con­stantly un­der siege and forced to en­gage in pub­lic re­la­tions (see quotes sec­tions L and M) and is con­stantly fac­ing Asym­met­ric Jus­tice and the Copen­hagen In­ter­pre­ta­tion of Ethics. This puts tremen­dous pres­sure on cor­po­ra­tions to tell differ­ent sto­ries to differ­ent au­di­ences, to avoid cre­at­ing records, and oth­er­wise en­gage in the types of be­hav­ior that will be com­fortable to the in­fected and un­com­fortable to the un­in­fected.

Another ex­pla­na­tion is that those who are in­fected don’t only re­ward each other within a cor­po­ra­tion. They also do busi­ness with and co­op­er­ate with the in­fected el­se­where. In­fected peo­ple are com­fortable with oth­ers who are in­fected, and un­com­fortable with those not in­fected, be­cause if the time comes to play ball, they might re­fuse. So those who re­fuse to play by these rules do bet­ter at ob­ject-level tasks, but face al­li­ances and hos­tile ac­tion from all sides, in­clud­ing cap­i­tal mar­kets, com­peti­tors and gov­ern­ment, all of which are, to vary­ing de­grees, in­fected.

I am likely miss­ing ad­di­tional mechanisms, ei­ther be­cause I don’t know about them or for­got to men­tion them, but I con­sider what I see here suffi­cient. I am no longer con­fused about short ter­mism.