How Doomed are Large Organizations?

Link post

We now take the model from the pre­vi­ous post, and ask the ques­tions over the next sev­eral posts. This first an­swer post asks these ques­tions:

  1. Are these dy­nam­ics the in­evitable re­sults of large or­ga­ni­za­tions?

  2. How can we fore­stall these dy­nam­ics within an or­ga­ni­za­tion?

  3. To what ex­tent should we avoid cre­at­ing large or­ga­ni­za­tions?

  4. Has this dy­namic ever been differ­ent in the past in other times and places?

Th­ese are the best an­swers I was able to come up with. Some of this is re­it­er­a­tion of pre­vi­ous ob­ser­va­tions and pre­scrip­tions. Some of it is new.

There are some bold claims in these an­swer posts, which I lack the space and time to defend in de­tail or provide cita­tions for prop­erly, with which I am con­fi­dent many read­ers will dis­agree. I am fine with that. I do not in­tend to defend them fur­ther un­less I see an op­por­tu­nity in do­ing so.

I would love to be miss­ing much bet­ter strate­gies for mak­ing or­ga­ni­za­tions less doomed – if you have ideas please please please share them in the com­ments and/​or el­se­where.

Are these dy­nam­ics the in­evitable re­sult of large or­ga­ni­za­tions?

Th­ese dy­nam­ics are the de­fault re­sult of large or­ga­ni­za­tions. There is con­tin­u­ous pres­sure over time push­ing to­wards such out­comes.

The larger the or­ga­ni­za­tion, the longer it ex­ists, and the more such out­comes have already hap­pened, both there and el­se­where, the greater the pres­sure to­wards such out­comes.

Once such dy­nam­ics take hold, re­vers­ing them within an or­ga­ni­za­tion is ex­tremely difficult.

Non-lo­cally within a civ­i­liza­tion, one can al­low new or­ga­ni­za­tions to pe­ri­od­i­cally take the place of old ones to re­set the dam­age.

Lo­cally within a suffi­ciently large or­ga­ni­za­tion and over a suffi­ciently long time hori­zon, this makes these dy­nam­ics in­evitable. The speed at which this oc­curs still varies greatly, and de­pends on choices made.

How can we fore­stall these dy­nam­ics within an or­ga­ni­za­tion?

Th­ese dy­nam­ics can be fore­stalled some­what through a strong or­ga­ni­za­tional cul­ture that de­votes sub­stan­tial head space and re­sources to keep­ing the wrong peo­ple and be­hav­iors out. This re­quires a leader who be­lieves in this and in mak­ing it a top pri­or­ity. Usu­ally this per­son is a founder. Los­ing the founder is of­ten the trig­ger for a rapid ramp up in maze level.

Keep­ing maze lev­els in check means con­tin­u­ously sac­ri­fic­ing sub­stan­tial head space, re­sources, abil­ity to scale and short-term effec­tive­ness to this cause. This holds both for the or­ga­ni­za­tion over­all and the leader per­son­ally.

Head space is sac­ri­ficed three ways: You have less peo­ple, you de­vote some of those peo­ple to the maze-fight­ing pro­cess, and the pro­cess takes up space in ev­ery­one’s head.

Cen­tral to this is to ruth­lessly en­force an or­ga­ni­za­tional cul­ture with zero tol­er­ance for maze be­hav­iors.

Do­ing any­thing with an in­tent to de­ceive, or an in­tent to game your met­rics at the ex­pense of ob­ject level re­sults, needs to be an au­to­matic “you’re fired.”

Some amount of poli­tics is a hu­man uni­ver­sal, but it needs to be strongly dis­cour­aged. Similarly, some amount of putting in ex­tra effort at cru­cial times is nec­es­sary, but strong pat­terns of guard­ing peo­ple’s non-work lives from work, both in terms of time and other in­fluences, are also strongly nec­es­sary.

Work­ers and man­agers need to have as much effec­tive skin in the game as you can muster.

One must hire care­fully, with a keen eye to the mo­ti­va­tions and in­stincts of ap­pli­cants, and a long pe­riod of teach­ing them the new cul­tural norms. This means at least grow­ing slowly, so new peo­ple can be prop­erly in­cor­po­rated.

You also want a rel­a­tively flat hi­er­ar­chy, to the ex­tent pos­si­ble.

There will always be bosses when crunch time comes. Some­one is always in charge. Don’t let any­one tell you differ­ent. But the less this is felt in or­di­nary in­ter­ac­tions, and thus the more tech­ni­cally di­rect re­ports each boss can have and still be effec­tive, and thus the less lev­els of hi­er­ar­chy you need for a given num­ber of peo­ple, the bet­ter off you’ll be.

You can run things in these ways. I have seen it. It helps. A lot.

Another ap­proach is to lower the out­side maze level. Do­ing so by chang­ing so­ciety at large is ex­ceed­ingly hard. Do­ing so by as­so­ci­at­ing with out­side or­ga­ni­za­tions with lower maze lev­els, and go­ing into in­dus­tries and prob­lems with lower maze lev­els, seems more re­al­is­tic. If you want to ‘dis­rupt’ an area that is suffer­ing from maze dys­func­tion, it makes sense to by­pass the ex­ist­ing sys­tems en­tirely. Thus, move fast, break things.

One can think of all these tac­tics as tak­ing the ques­tions one uses to iden­tify or pre­dict a maze, and try­ing to en­g­ineer the an­swers you want. That is a fine in­tu­itive place to start.

How­ever, if Good­hart’s Law alarm bells did not go off in your head when you read that last para­graph, you do not ap­pre­ci­ate how dan­ger­ous Good­hart Traps are.

The Good­hart Trap

The fatal flaw is that no mat­ter what you tar­get when dis­tribut­ing re­wards and pun­ish­ments and cul­tural ap­proval, it has to be some­thing. If you spell it out, and a suffi­ciently large or­ga­ni­za­tion has lit­tle choice but to spell it out, you in­evitably re­place one type of Good­hart­ing with an­other. One type of de­cep­tion be­comes an­other.

One uni­ver­sal is that in or­der to main­tain a unique cul­ture, you must filter for those that hap­pily em­brace that cul­ture. That means you are now test­ing ev­ery­one con­stantly, no mat­ter how ex­plicit you avoid mak­ing this, on whether they hap­pily em­brace the com­pany and its cul­ture. Peo­ple there­fore pre­tend to em­brace the cul­ture and pre­tend to be con­stantly happy. Even if they do em­brace the cul­ture and are happy, they still ad­di­tion­ally will put on a show of do­ing so.

If you pun­ish de­cep­tion you get peo­ple pre­tend­ing not to de­ceive. If you pun­ish pre­tend­ing, you get peo­ple who pre­tend to not be the type of peo­ple who would pre­tend. Peo­ple Good­hart on not ap­pear­ing to Good­hart.

Which is a much more in­ter­est­ing level to play on, and usu­ally far less de­struc­tive. If you do a good enough job pick­ing your Good­hart tar­gets, this beats the al­ter­na­tives by a lot.

Still, you even­tu­ally end up in a ver­sion of the same place. De­cep­tion is de­cep­tion. Pre­tend­ing is pre­tend­ing. Fraud is fraud. The soul still dies. Si­mu­lacrum lev­els still slowly rise.

Either you strongly en­force a cul­ture, and slowly get that re­sult, or you don’t. If you don’t and are big enough, you quickly get a maze. If you do and/​or are smaller, de­pend­ing on your skill level and ded­i­ca­tion to the task, you slowly get a maze.

Hiring well is bet­ter than en­forc­ing or train­ing later, since once peo­ple are in they can then be them­selves. Also be­cause en­force­ment of cul­ture is, as pointed out above, toxic even if you mean to en­force a non-toxic ideal. But rely­ing on em­ployee se­lec­tion puts a huge pre­mium on not mak­ing hiring mis­takes. Even one bad hire in the wrong place can be fatal. Espe­cially if they then are in a po­si­tion to bring oth­ers with them. You need to defend your hiring pro­cess es­pe­cially strongly from these same cor­rup­tions.

My guess is that once an or­ga­ni­za­tion grows be­yond about Dun­bar’s num­ber, defend­ing your cul­ture be­comes a los­ing bat­tle even un­der the best of cir­cum­stances. En­forc­ing the cul­ture will fail out­right in the medium term, un­less the cul­ture out­side the or­ga­ni­za­tion is sup­port­ing you.

If you are too big, ev­ery known strat­egy is only a hold­ing ac­tion. There is no per­ma­nent solu­tion.

To what ex­tent should we avoid cre­at­ing large or­ga­ni­za­tions?

Quite a lot. Th­ese effects are a re­ally big deal. Or­ga­ni­za­tions get less effec­tive, more toxic and cor­rupt as places to work and in­ter­act with, and add more tox­i­c­ity and cor­rup­tion to so­ciety.

Every level of hi­er­ar­chy en­hances this effect. The first five, dra­mat­i­cally so. Think hard be­fore be­ing or hav­ing a boss. Think harder be­fore let­ting some­one’s boss re­port to a boss. Think even harder than that be­fore adding a fourth or fifth level of hi­er­ar­chy.

That does not mean such things can be fully avoided. The ad­van­tages of large or­ga­ni­za­tions with many de­grees of hi­er­ar­chy are also a re­ally big deal. We can­not avoid them en­tirely.

We must treat cre­at­ing ad­di­tional man­age­rial lev­els as hav­ing very high costs. This is not an ac­tion to be taken lightly. Wher­ever pos­si­ble, cre­ate dis­tinct or­ga­ni­za­tions and al­low them to in­ter­act. Even bet­ter, al­low peo­ple to in­ter­act as in­di­vi­d­u­als.

This adds fric­tion and trans­ac­tion costs. It makes many forms of co­or­di­na­tion harder. Some­times it sim­ply can­not be done if you want to do the thing you’d like to do.

This is in­creas­ingly the case, largely as a re­sult of en­emy ac­tion. Some of this is tech­nol­ogy and our prob­lems be­ing le­gi­t­i­mately more com­plex. Most of it is reg­u­la­tory frame­works and maze-sup­port­ing so­cial norms that re­quire mas­sive costs, in­clud­ing mas­sive fixed costs, be paid as part of do­ing any­thing at all. This is a key way mazes ex­pro­pri­ate re­sources and re­ward other mazes while pun­ish­ing non-mazes.

I of­ten ob­serve peo­ple who are stuck work­ing in mazes who would much pre­fer to be self-em­ployed or to exit their cur­rent job or lo­ca­tion, but who are un­able to do so be­cause the le­gal deck is in­creas­ingly stacked against that.

Even if the work it­self is per­mit­ted, health in­surance is­sues alone force many into work­ing for the man.

When one has a suc­cess­ful small or­ga­ni­za­tion, the nat­u­ral in­stinct is to scale it up and be­come a larger or­ga­ni­za­tion.

Re­sist this urge when­ever pos­si­ble. There is noth­ing wrong with be­ing good at what you do at the scale you are good at do­ing it. Set an ex­am­ple oth­ers can em­u­late. Let oth­ers do other things, be other places. Any prof­its from that en­ter­prise can be re­turned to in­vestors and/​or paid to em­ploy­ees, and used to live life or cre­ate or in­vest in other pro­jects, or to help oth­ers.

One need not point to ex­plicit quan­tified dan­gers to do this. Ar­gu­ments that one can­not le­gi­t­i­mately choose to ‘leave money on the table’ or oth­er­wise not max­i­mize, are max­i­mal­ist ar­gu­ments for some util­ity func­tion that does not prop­erly cap­ture hu­man value and is sub­ject to Good­hart’s Law, and against the le­gi­t­i­macy of slack.

The fear that if you don’t grow, you’ll get ‘beaten’ by those that do, as in Ray­mond’s king­doms? Overblown. Also ask­ing the wrong ques­tion. So what if some­one else is big­ger or more su­perfi­cially suc­cess­ful? So what if you do not build a gi­ant thing that lasts? Every­thing ends. That is not, by de­fault, what mat­ters. A larger com­pany is of­ten not bet­ter than sev­eral smaller com­pa­nies. A larger club is of­ten not bet­ter than sev­eral smaller clubs. A larger state is of­ten not bet­ter or longer last­ing than sev­eral smaller ones. Have some­thing good and pos­i­tive, for as long as it is vi­able and makes sense, rather than trans­form­ing into some­thing likely to be bad.

Peo­ple like to build em­pires. Those with power usu­ally want more power. That does not make more power a good idea. It is only a good idea where it is in­stru­men­tally use­ful.

In some places, com­pe­ti­tion re­ally is win­ner-take-all and/​or reg­u­la­tions and con­di­tions too heav­ily fa­vor the large over the small. One must grow to sur­vive. Once again, we should be sus­pi­cious that this dy­namic has been en­g­ineered rather than be­ing in­her­ent in the un­der­ly­ing prob­lem space.

Espe­cially in those cases, this leads back to the ques­tion of how we can grow larger and keep these dy­nam­ics in check.

Has this dy­namic ever been differ­ent in the past in other places and times?

Th­ese dy­nam­ics seem to me to be get­ting in­creas­ingly worse, which im­plies they have been bet­ter in the past.

Re­cent de­vel­op­ments in­di­cate an in­creas­ing simu­lacrum level, an in­creas­ing re­luc­tance to al­low older in­sti­tu­tions to be re­placed by newer ones, and an in­creas­ing re­li­ance on crony­ism and cor­rup­tion that props up failure, al­low­ing mazes to sur­vive past when they are no longer able to fulfill their origi­nal func­tions.

Those in the poli­ti­cal and aca­demic sys­tems, on all sides, in­creas­ingly openly ad­vo­cate against the very con­cept of ob­jec­tive truth, or that peo­ple should tell it, or are blame­wor­thy for not do­ing so. Our pres­i­dent’s sup­port­ers ad­mit and ad­mire that he is a cor­rupt liar, claiming that his hon­esty about his cor­rup­tion and ly­ing, and his ad­mira­tion for oth­ers who are cor­rupt, who lie and who bully, is re­fresh­ing, be­cause they are dis­tinct from the cor­rupt, the liars and the bul­lies who are more lo­cally rele­vant to their lives. Dis­course is in­creas­ingly fraught and difficult. When some­one wants to en­gage in dis­course, I fre­quently now ob­serve them spend­ing much of their time point­ing out how difficult it is to en­gage in dis­course (and I am not claiming my­self as an ex­cep­tion here), as op­posed to what such peo­ple used to do in­stead, which was en­gage in dis­course.

We are in­creas­ingly par­a­lyzed and un­able to do things across a wide va­ri­ety of po­ten­tial hu­man ac­tivi­ties.

Ex­pro­pri­a­tion by ex­ist­ing mazes and sys­tems eats in­creas­ing shares of ev­ery­thing, es­pe­cially in ed­u­ca­tion, health care and hous­ing.

I don’t have time for a full take­down here, but: Claims to the con­trary, such as those re­cently made by Alex Tab­brok in Why Are The Prices So Damn High?, are statis­ti­cal ar­ti­facts that defy the ev­i­dence of one’s eyes. They are the product of Moloch’s Army. When I have in­surance and am asked with no warn­ing to pay $850 for liter­ally five min­utes of a doc­tor’s time, af­ter be­ing kept wait­ing for an hour (and ev­ery­one I ask about this says just re­fuse to pay it)? When send­ing my child to a kinder­garten costs the ma­jor­ity of a skil­led ed­u­ca­tor’s salary? When you look at rents?

Don’t tell me the prob­lem is la­bor costs due to in­creas­ing de­mand for real ser­vices.

Just. Don’t.

Some tech­nolog­i­cal in­no­va­tions re­main per­mit­ted for now, and many of the or­ga­ni­za­tions ex­ploit­ing this are rel­a­tively new and re­li­ant on ob­ject-level work, and thus less maze-like for now, but this is suffi­ciently nar­row that we call the re­sult “the tech in­dus­try.” We see rapid progress in the few places where in­no­va­tion and ac­tual work is per­mit­ted to those with­out mazes and con­nec­tions, and where there is suffi­cient mo­ti­va­tion for work, ei­ther in­trin­sic or mon­e­tary.

The tech in­dus­try also ex­hibits some very maze-like be­hav­iors of its own, but it takes a differ­ent form. I am un­likely to be the best per­son to tackle those de­tails, as oth­ers have bet­ter di­rect ex­pe­rience, and I will not at­tempt to tackle them here and now.

We see very lit­tle ev­ery­where else. In­creas­ingly we live in an amalga­mated gi­ant maze, and the maze is par­a­lyz­ing us and tak­ing away our abil­ity to think or talk while rob­bing us blind. Mazes are in­creas­ingly in di­rect po­si­tion to cen­sor, de­plat­form or pun­ish us, even if we do not work for them.

The idea of pos­i­tive-sum, ob­ject-level in­ter­ac­tions be­ing some­one’s pri­mary source of in­come is in­creas­ingly seen as ille­gi­t­i­mate, and risky and ir­re­spon­si­ble, in con­trast to work­ing for a maze. Peo­ple in­stinc­tively think there’s some­thing shady or re­bel­lious about that en­tire en­ter­prise of hav­ing an ac­tual en­ter­prise. A proper per­son seeks rent, plays the game start­ing in child­hood, sends the right sig­nals and finds ways to game the sys­tem. They in­crease their ap­peal to mazes by mak­ing them­selves as de­pen­dent on them and their in­come and le­gi­t­i­macy streams, and as vuln­er­a­ble to their black­mail, as pos­si­ble.

The best way to see that pos­i­tive-sum games are a thing is to no­tice that the sum changes. If ev­ery­thing is zero-sum, the sum would always be zero.

The best way to see that these dy­nam­ics used to be much less se­vere, at least in many times and places, is that those times and places looked and felt differ­ent, and got us here with­out col­laps­ing. Mo­ral Mazes was writ­ten be­fore I was born, but the spread of these dy­nam­ics is clear as day within my life­time, and yours as well.

Did some times and places, in­clud­ing our re­cent past, have it less bad than us in these ways? I see this as al­most cer­tainly true, but I am un­cer­tain of the mag­ni­tude of this effect due to not hav­ing good enough mod­els of the past.

Did some times and places have it worse than we do now? Very pos­si­ble. But they’re not around any­more. Which is how it works.

The next sec­tion will ask why it was differ­ent in the past, what the causes are in gen­eral, and whether we can du­pli­cate past con­di­tions in good ways.