John Wentworth on bureaucracy:

Responding to Chris: if you go look at real bureaucracies, it is not really the case that "at each level the bosses tell the subordinates what to do and they just have to do it". At every bureaucracy I've worked in/around, lower-level decision makers had many de facto degrees of freedom. You can think of this as a generalization of one of the central problems of jurisprudence: in practice, human "bosses" (or legislatures, in the jurisprudence case) are not able to give instructions which unambiguously specify what to do in all the crazy situations which come up in practice. Nor do people at the top have anywhere near the bandwidth needed to decide every ambiguous case themselves; there is far too much ambiguity in the world. So, in practice, lower-level people (i.e. judges at various levels) necessarily make many many judgement calls in the course of their work.

Also, in general, tons of information flows back up the hierarchy for higher-level people to make decisions. There are already bureacracies whose purpose is very similar to HCH: they exist to support the decision-making of the person at the top. (Government intelligence is a good example.) To my knowledge/experience, such HCH-like bureacracies are not any less dysfunctional than others, nor do normal bureacracies behave less dysfunctionally than normal when passing information up to a high-level decision maker.

Responding to Joe: if you go look at real bureaucracies, most people working in them are generally well-meaning and trying to help. There is still a sense in which incentives are a limiting factor: good incentives are information-carriers in their own right (like e.g. prices), and I'll link below to arguments that information-transmission is the problem. But incentives are not the problem in a way which can be fixed just by having everyone share some non-selfish values.

So why do bureaucracies (and large organizations more generally) fail so badly?

My main model for this is that interfaces are a scarce resource. Or, to phrase it in a way more obviously relevant to factorization: it is empirically hard for humans to find good factorizations of problems which have not already been found. Interfaces which neatly split problems are not an abundant resource (at least relative to humans' abilities to find/build such interfaces). If you can solve that problem well, robustly and at scale, then there's an awful lot of money to be made.

Also, one major sub-bottleneck (though not the only sub-bottleneck) of interface scarcity is that it's hard to tell who has done a good job on a domain-specific problem/question without already having some domain-specific background knowledge. This also applies at a more "micro" level: it's hard to tell whose answers are best without knowing lots of context oneself.

I should also mention: these models came out of me working in/around bureacratic organizations, as they were trying to scale up. I wanted to generally understand the causes of various specific instances of dysfunction. So they are based largely on first-hand knowledge.

I can't find the clip, but the excellent satire show Yes, Minister suggests the theory that bureaucracies can be thought of as budget-maximizers. The quote was something like "In private industry, profit is the measure of your success. In public service, there are no profits, so instead your importance is measured by how many people and how much money you control."

Bureaucracies exist inside larger firms as well, and share a lot of elements with public bureaucracies.  Much of the "this is stupid, but nobody knows how to make it better" (aka "bwuh?") feeling applies equally.

I think a reasonable model for it is "mission motive" - somewhat like any other motive, but with a very weak or missing feedback mechanism.  Without being able to track results, and with no market discipline (failure -> bankruptcy when the motive is aligned with existence), you get weird behaviors based on individual humans with unrefined models.

Curated. 

I've read a bunch of posts being confused or annoyed at bureaucracies, and I found this post a helpful wake-up call and Open Problem Statement: "why don't we have any simple models that adequately explain what bureaucracies do?"

I don't know that I believe there will turn out to such a simple model or unified-theory-of-bureaucracy, but I'd expect the search for one to be worthwhile and churn up interesting things along the way and hopefully output a collection of models that get more useful and easy-to-apply over time.

Something about this also makes think of True Names as a tool to combat goodhart. Bureaucracies are an instance of the generalized alignment problem, and having a good model of what sort of incentives your creating can help you use bureaucracies on purpose.

So I'm asking you. What might be the quantity whose maximization or minimization is the basic general explanation of what bureaucracies do?

Let me try my hand at sketching a theory.

Here's some background first. Firms are well-modeled as profit maximizers because, although they employ internal bureaucracies to achieve their ends, bureaucracies that are bad at the task of making investors money are either selected out over time due to competition, or are pruned due to higher-up managers having relatively strong incentives to fire people who are not making investors money. This model relies on an assumption that investors themselves are usually profit-maximizers, which seems uncontroversial. 

By contrast, government bureaucracies lack the pressures of competitition, though they can (though less commonly) be subject to pruning, especially at the higher levels. I can think of two big forces shaping the motivations of government bureaucracies: the first being internal pressures on workers to "do work that looks good" to get promoted, and the second being a pressure to conform to the desires of the current president's political agenda (for people at the top of the bureaucracy).

The second pressure mainly acts through a selection effect: the president generally nominates someone to head the CDC, FDA etc. They select this person on the basis of how well they can achieve their broad agenda, which is subject to political constraints. Bryan Caplan's work on voter theory and irrationality is the primary workhorse in this part of the theory. 

Since arguably the main way a president acts in their role as a politician is by nominating people to lead bureaucracies, we should expect them to take special care in ensuring that any potential nominee will conform to the values/biases of the general public. These values/biases include all the ones that Bryan Caplan mentions, plus a few more that are relevant for our purposes: pro "doing something" even if there's not much we can do, and pro doing "feel-good" measures.

The first pressure is where the blame-minimization theory plays its largest role. Each person in the chain doesn't want to be blamed for a big disaster, and the worst thing any one of them could do is make some maverick judgement completely out of line with conventional wisdom in the field, only for it to blow up in their face. Thus, we should expect a highly conformist, risk-averse culture within bureaucracies, focused on doing what looks and sounds good to please internal evaluators, at the cost of doing what might in fact be good.

This story doesn't lend itself to a simple theory like "maximize profit" or "maximize the probability of getting a majority of votes" does. However, we can at least make the following predictions:

• If public opinion contradicts expert consensus, government bureaucracies will generally side with public opinion, not expert consensus. Given the strong constraints on the higher-up leadership to please the president, acting against public opinion is essentially a death-wise for any would-be head of the agency. There might be some especially strong internal pressure to side with expert consensus at the cost of losing political points, but these pressures will be selected out over time as people realize it's more comfortable to simply follow public opinion.
• If there's a short-term internal performance metric that is broadly associated with long-term performance, but far easier to measure than actual long-term success, we should expect people within bureaucracies to maximize this internal performance metric, even in cases where the association breaks down. This is just another way of saying that government bureaucracies Goodhart performance, in contrast to private corporations, which have much stronger incentives to make money even if that requires doing something anti-conformist.

Assuming the bureaucracy is hierarchical, the maximizer may vary depending on the level. At the lower levels, a process-maximizer may best model behavior. Map versus territory. Akin to a mis-aligned AI paperclip-maximizer, reward is based on adherence to process, results do not matter. Mid-hierarchical levels are budget-maximizers. Body-count may be a surrogate. The bureaucratic topology that emerges and morphs at these mid-levels is where things become chaotic for the higher levels. Perhaps entrenchment, power, consensus, and hubris-maximizers join the dance. Predicting behavior at these higher levels may be more a matter of profiling than modeling. Regardless, the bureaucracy as a whole is more like an oil tanker than a jet ski. Its behavior in the near term is rather obvious.

I think there's a conflation happening here, turning bureaucracy into a noun and then overfitting it to all instances of bureaucracy.  There are highly bureaucratic teams/organizations among the Fortune 500, and there are more agile government agencies.

1.
I contend that bureaucracy develops to minimize downside risk.  Organizations create rules, processes and structures with the goal of making sure tasks are constantly done the same way, and importantly, limiting (ideally preventing) the chance of something going wrong (e.g., missing critical data on a form, not looping in all of the appropriate people.)
 

1.a
What I find most interesting is the trade-off that this implies... de-risking/limiting downside potential almost tautologically implies a restriction on upside potential. Organizations may be willing to get less out-performance from employees, in exchange for reducing the "harm" an employee could do... which is very stifling for many high-potential individuals.

When you stop and think about it so many jobs, both entry-level and management, are simply executing or overseeing some web of repeated tasks, that enable or interact with another web of repeated tasks in a different part of the organization.

2.
A separate point is that many of the organizations we refer to as bureaucracies are government agencies, so the interesting question is whether bureaucracy is more prevalent in those types of organizations.

Certainly, there's an easy narrative about not having a market to discipline them, so governmental agencies are likely to outlive their usefulness and become sclerotic.

However, I think there's also an observation bias; the media exists in no small part to shine a light on government waste/excess. What's more, they are generally funded by taxpayer dollars and most people have semi-regular interactions with many agencies. But ask most people how bureaucratic their employers are and you'll likely get a tirade about HR and Operations and TPS reports that nobody reads.  The reality is, most of us don't really consider the nature of private bureaucracies on a regular basis, and if we do, it's usually constrained to the ones we have some personal interaction wit

2.a
In general, the larger the organization, the more bureaucracy you expect to find; it's part of the "infrastructure" holding the organization together! So the interesting question here is what is the optimal amount of bureaucracy given the size/nature of your organization/project.

2.b
In the long term, the point about market forces eliminating bureaucracy is probably true, with at least one caveat. Competition means that young, nimble companies are likely to eliminate stagnant, bureaucratic companies, though it may take many years for this to happen.

The caveat is regulation: government may stay in and set up standards and requirements that impose de facto bureaucracy. Beyond this direct introduction of bureaucracy (and the second-order bureaucracy meant to manage the requirements the regulation put in place), it also tends to make markets less competitive (a young startup likely can't bear the burden of the regulation and so either doesn't enter the market or has to find a clever way to circumnavigate them (not always a positive.) Regulation, therefore, is a form of bureaucracy and bears the same trade-offs; you give up some upside to limit some downside.

3. An interesting topic to think about is when is regulation/bureaucracy appropriate? (Warning:  wandering into libertarian think-space here.) These are usually put in place to limit specific, known/anticipated downsides, but at the cost of largely unknown upsides. Of course, maybe they also limit unknown downsides. The unknown unknowns are challenging to handle... the right answer is likely to involve quick but limited/targeted regulation/bureaucracy. So there's even more room to discuss how exactly to set that up.

And so it goes...

Bureaucracies create jobs. For mechanistic details see Parkinson's Law.

Regarding moving beyond blame minimization, I think it's worth mentioning my Venture Granters, a system for protecting sane risk-takers in public funding institutions: https://www.lesswrong.com/posts/NY9nfKQwejaghEExh/venture-granters-the-vcs-of-public-goods-incentivizing-good

I love that this post is asking a important and big question at a pleasing level of abstraction, aiming for something productive. 

(So as not to bury the lede: by the end, I will have gotten around to arguing that bureacracies are adaptation executors, where the adaptations are laws that can be good or bad.)

After reading this lovely post I hit ^fcapture and was mildly surprised that no one mentioned regulatory capture.

If you try ^fcorruption you easily find Dumbledore's Army's comment which is highly upvoted and somewhat pessimistic. 

I'm going to try to sketch in more things along a similar like, with more standardized keywords, and more links, to weave together abstract ideas and concrete examples of those ideas.

As mentioned in other comments, a bureaucrat can play "do almost nothing but the procedural minimum" and collect checks. 

They can try to conspire to rise using Moral Maze tactics against each other, playing a game whose broad strategic outlines have existed for a long time which can be called patrimonial bureacracy. 

Alternatives to patrimonial bureaucracy are kind of hard to talk about, maybe like the old saw about how fish might not have a word for "wetness", but my current working theory here is that a contrast object exists in the form of jurisprudential bureacracy (which almost no one talks about, and the google hit to that master's thesis which finally uses the phrase on page 53 is typically narrow and obscure). 

The gist is: there used to be noblemen who had inherited power to tax peasants. They semi-voluntarily joined more and more formal sorts of feudal factional alliances. They implicitly took for granted the right to brutalize and steal from THEIR OWN peasants in certain circumstances, but they had to justify such actions of with the peasants of their feudal allies. 

The upshot was an environment where the noble-turned-bureacrats would argue in terms of the general principles which naturally apply to the general rights of feudal lords and their duties to each other and (minimally) to their peasants. Especially in central Europe (ancient germany?) "kings" were often de facto so weak as to be unable to rule except by the non-trivial consent of lesser nobles. 

So clear norms and clear reasoning within jurisprudential bureaucracy were possibly important in sort of the way (and probably for very similar reasons) that gave ocean-going sea pirates non-trivial loot sharing and governance norms?

In this model: pirates and feudal lords might actually be MORE "morally proper seeming" than patrimonial bureaucrats because at least they had meaningful principles? Also they had some agency. (But the expectation was that their agency and principles would be used to prey upon peasants or defenseless boats or whatever.)

EXCEPT: what if the King at the top of the patrimonial system was Causing the pirate/bureaucrats to NOT necessarily follow norms, but actually just to to serve the King's whims? 

And what if the King was benevolent and had "good whims"?

HOWEVER, the very top dog of the bureaucracy, in nominally-modern nominally-democratic nominally-rights-respecting governments is supposed to be an elected official who could function kind of like a "Strong King" or a "actually somehow powerful head pirate" whose power over the other parts of the government descends from the consent of the median-voter. 

So maybe a patrimonial bureaucracy with a Good president could be Good? 

It kinda stands to reason that it would work this way? :-)

Note... the median voter would be patrimonially governed by a bureaucracy that is patrimonially governed by an elected official (who is electorally governed by the median-voter... (etc))) and this pattern makes it (hopefully) such that the elected leader could hypothetically impose principles or at least "know which side your bread is buttered on incentives" to the bureaucrats in the bureaucracy. 

In the US, this is how it worked for a while.

HOWEVER: much of this logic was dismantled, with all such oversight deemed "political patronage", in the 1883 Pendleton Reform.

(Arguably this actually was an objective and huge improvement in governance, but part of this might be that it rebooted an old system that could be goodhearted (by electoral crony systems that hacked voters to gain lots of government jobs) with a new system that could hypothetically be goodhearted but only by less mature and mesa-optimizers? ....and then maybe the post-Pendleton system WAS getting goodhearted and starting to rot by the 1920s and 1930s... 

...but then FDR came in and did a sort of Good(?) Powerful King style reboot (incidentally inventing the NSF and assembling the CIA and a huge number of other now-normal and now-natural seeming institutions) which might have been terrible according to some abstract standards but had the practical virtue of: winning World War 2 for the US... And if you're going to pick TWO virtues, then maybe ONE of them should probably be "being able to win giant wars"?

So this perhaps provides history and theory to power a deep claim here that "patrimonial bureaucracies don't natively actually coherently maximize anything" but rather they might just start as the (metaphorical) adaptation executors of that someone, like perhaps a visionary goal-seeking founders, just wants to happen. Do X, do Y, do Z, and later someone with a brain can look at XYZ+results to see if the system works or needs to change or whatever. (Hat tip here to Gunnar who mentioned relevant things in a neighboring comment before me and other hat tip to Samo himself).

HOWEVER, I think that over time patrimonial bureaucracies probably devolve into "governance cancer"?

Maybe? Probably? Sometimes? Eventually always?

This is my own not-entirely-certain-or-well-supported hunch. If it could even be operationalized and researched I bet the diagnosis of this or that bureaucracy as "having already metastasized by date D" would be controversial... but if that was possible then I think it would be clearly "benevolently high value long term institution design research" to know what founding bureaucratic features cause the longest amount of time between (hopefully) BENEVOLENT bureaucratic founding and KLEPTOCRATIC bureaucratic metastasis.

There's a deep analogy lurking here, I think, between bureaucracies, cancer, and mesa-optimization.

Part of why you kind of should predict actual human bureaucrats in actual institutions to eventually begin to be unaligned mesa-optimizers within a bureaucracy is a thing that Dumbedore's Army might have missed...

...a key way for bureaucrats to gain value is by using their position in the bureaucracy to build allies and the PIVOTING at some point to working OUTSIDE of the bureaucracy for a lot of money. 

This is called the revolving door problem when human government people do it. I haven't yet found (but also haven't looked super hard) for equivalent problems in biology. The closest biological analogy I have so far is prophages where "lysis" is kind of like "resign and work as a lobbyist".

For example Scott Gottlieb was the head of the FDA and a mere 8 months before covid started in Wuhan he resigned from his role as the head of the FDA (whose legislative mandate is objectively dumb, and this dumbness made the covid response much much worse). Then about 5 months after covid (so this would be maybe 2 months after Moderna was testing the mRNA vaccine but still while the medical prototypes were illegal to sell no matter what opt-outs clever people might want to try) Scott Gottlieb joined the board of directors of Pfizer. 

Presumably his job was to "give them insight" into "the process of regulatory compliance"?

If we had a coherent theory of public service, this might be illegal? (Certainly if it WAS illegal then joining the government would basically be a lifelong commitment, and the pensions and lack of fear of firing or layoffs would make a lot of sense.)

Something I've noticed is that often people THINK lots of sort of obviously bad things like regulators working right afterwards for the people they regulate IS ILLEGAL because they assume the government runs in a morally sane way, and then are unhappily surprised when there is no law against such things.

Caption: the penny just dropped for him, when he realized how much is NOT illegal, and already happening.

Grant that the US government won't actually provide the public good of "adequate infectious disease fighting", a way for private entities to make a profit by providing private goods to private purchasers RELIES on this large scale structural failure to provide an obvious public good is: to make and sell expensive individual medical coping mechanisms (like masks and imperfect vaccines and so on) to each patient, so each patient can mitigate the effect of "the public goods disaster" for themselves, by themselves, for their own benefit, in a personally "validly selfishly" way.

This is maybe not bad? If Pfizer was making money off of a situation that the government is tragically "merely incapable" of fixing that would be one thing. But hiring former government regulators is an extra layer of: WTF?

Also, the regulatory bureacrats and politicians in New Zealand and Taiwan and Singapore and China proved that "it can be done" pretty independently of being a democracy or an authoritarian dictatorship or whatever. 

Solving covid was possible. Therefore the US's failed half-attempt to solve covid is less excusable.

And so that "general failure" is perhaps part of how Scott Gottlieb's new bureaucracy (the one that runs for-profit and has him as a board of director) remains happy in the status quo.

Also, they are almost certainly a beneficiary of regulatory capture by the FDA in a more direct way that has been true for decades, but harder to see because there wasn't a novel plague: the big companies capture value instead of generating consumer surplus... as a side effect of not having to compete with the (currently non-existent because illegal-at-low-prices) cheaper competitors who are less able to jump all the sometimes-situationally-pointless FDA hurdles whose existence: (1) lets Pfizer have many fewer competitors and (2) therefore have substantial pricing power.

Like, with competition like we have for food, the margins on medical manufacturing could be almost zero... and maybe that would be a good thing.  

If it happened, then I personally would be in favor of some kind of drug manufacturing subsidies (like the US has for food) to avoid gluts and shortages of medicine.

There's a strong tendency, at this juncture, based on this kind of reasoning, to want to find alive and active PEOPLE responsible for the "bad personal behaviors and their macroscale consequences".

I try to resist this impulse and look for knobs that can be turned that are formal and logical. It turns out: they exist!

Look at the naively obvious  tippy top of one these systems... then look even higher than the "mere Scott Gottliebs of the world", and see the actual presidents and legislators, and then look up even higher than that to see THE LAWS.  

We have the rule of law, after all... right? 

Then we could back chain from bad laws to elect people who will fix the bad laws. Then they might do that. Then the better laws should back-chain to bureaucracies that run according to those laws. Right?

Then we could hypothetically just reason directly about the obvious contents of laws, and the object level domain that the laws authorize the regulation of, and we can imagine what the laws would tend to cause if the laws were followed...

The existence of such actual literal laws is why I think it makes a lot of sense to think of these bureaucracies as essentially adaptation executors rather <anything>-maximizers.

Admittedly, from within the present system, the whole "rule of law + democracy" formula might be very hard to do, and to get the MEDIAN voters "riled up enough" it might be necessary to stoop to personal attacks on Official Scrapegoats as a sort of "rationally-calculated semi-tragic PR plan to educate the median voter in a way that the median voter will listen to"?

BUT FIRST: the actual laws! There is very little justification for smashing the bureaucrats who are following the laws unless you have a legislative proposal for what OTHER laws they could obey instead.

So: the FDA's has the power, in the law, to make all drugs illegal-to-sell-by-default unless the drug makers have privately spent huge amounts on highly Formalized Bureaucratic Science, in advance of any sales, and the law that empowers them to do this is the Kefauver-Harris amendment. 

If this law is repealed, they might stop actively hurting people by engaging in actively harmful behavior. 

If all the FDA did was proactively buy drugs (OTC and prescription, both?) and test them for honest labeling of the ingredients and possible contaminants... I think that would be FANTASTIC. 

That would generate a huge public good. That information would be helpful in many ways to many people, none of who could generate it themselves in a profitable way. It would be useful to funnel the things that were learned by active testing back into the supply chain and also useful to publish to help consumers demand better things and not demand more poisonous things with false dosing and ingredients and warning claims on the labels.

That kind of work might explain to me why the FDA could still sanely employ lots of people? Maybe? (But this is NOT WHAT THEY DO.)

Instead they... simply forbid medical competition unless certain kinds of very expensive cargo-culted scientific rituals have been performed, and also defend the monopoly pricing power of any company that performs the expensive rituals. The tens of thousands of employees oversee these cargo culted science rituals.

(I understand this characterization of FDA trials as cargo-culted "science" could be considered problematic, but actually, the place where the MOST empirical knowledge about clinical practice arises for really reals is IN THE CLINIC. The FDA trials aren't super terrible. Some of them replicate and some of them generalize. However, they are not adequate, and yet they are treated as Officially Adequate. Every General Practice Doctor in the US is engaged in clinical practice, and any time they prescribe a drug they are following a diagnostic hunch and performing an N=1 experiment on a patient who feels bad enough to want to pay to be that kind of clinical medical scientific subject. This is how huge amounts of progress in medicine in the first half of the 1900s actually happened.

But that ongoing clinical practice doesn't currently "count as science" so... that helps the monopolies remain monopolies? Also it prevents the original engine of innovation actually innovate.

And in the meantime, the ongoing clinical practice of prescribing approved drugs and watching what happens often turns up safety issues that lead to drug recalls. Every drug recall of an approved drug is maaaybe good "because safety"? (Or it could be just given some kind of warning on the label, and left in the toolset of doctors for more rare or cautious or "hail mary" uses.) 

So every drug recall of an approved drug ALSO is a sign that the FDA's mandated trials can't actually "scientifically prove" what it claims that the medical trials proved: the safety (or efficacy) of the drug. 

In this sense, the FDA's processes are a cargo-cult of science. They ignore how the real knowledge was really acquired, and pretend more certainty up front, and pretend to be shocked by later "totally unpredictable" discoveries. Hence I say: cargo cult.

Also like: the existence of lots of cargo (from certain practices) is exactly what is LACKED by the Melanesia religions originally called "cargo cults". They waved around magic sticks that looked like "antennas to talk to planes" and no planes landed and gave them food. We wave "science" around so: Where is the medical science cornucopia?

Where is the cure for death? Where is the cure for cancer? Where is the cure for covid? If we had robust "non-cargo cult" medical science practices, we would either have clear answers for why it was formally theoretically impossible to construct such "medical cargo"... or we would have the cargo. 

In practice: we have neither. We don't know that cures like this are impossible and we also don't have the cures.

...

Pulling back more generally, I'm not saying CIA delenda est. I'm not saying NSF delenda est. (Remember above, FDR founded these other bureaucracies and we still have them.)

Those "executable (law-empowered) institutional adaptations" might have operations I don't understand that well. 

I think to say one way or the other that you would have to study the goals, constraints, and means and ends, and then look at the legislative and executive authorizations, and how those authorizations work in practice, and then decide one way or the other? Maybe? 

But I have published in virology and tried to raise money for biotech startups, and read textbooks on the topic and I am saying: FDA delenda est.

Speaking to the larger point about bureaucracies in general: I do think that the "highest unelected/unappointed bureaucrat" is often de facto STRONGER in modern democracies than the elected and/or appointed officials who nominally regulate them.

The elected officials are beholden to the median voter, and if that voter changes his or her mind they are potentially VERY transitory. 

Any time an elected official is clearly threatening to an (evil?) bureaucracy, the conspiracy of top bureaucrats has the incentives and sometimes the power to influence the median voter (such as by media leaks) to get rid of the elected official FASTER than the official can legislate them out of doing bad things.

The critical thing is the asymmetry: the bureaucrat is insulated from the median voter, while also having write access to the evidence and ideas in the head of the median voter through journalists. Thus, in a deep sense, the bureaucrat with the ear of journalists and the backing of their patrimonial clients within a bureaucracy is often STRONGER THAN a legislator or president that hypothetically might want to change the legislative mandate under which the bureaucrat operates.

Elected executive hire/fire power might be faster than media leaks... but that power IS ILLEGAL in most cases.

I think this is a deep structural problem, and the solution would be: the honor and goodness and sense of duty of the highest bureaucrats? Either that or a massive awakening among the voters to the operation of bureaucracies? And so in some sense maybe we've almost looped all the way around back to chivalrous knights using personally owned power to do good deeds (or just killing a bunch of peasants) according to their own conscience?

Except it doesn't even work like that! And this again reveals how dangerous it is to focus on the mere people, rather than the formal laws that formally empower them in a system that runs according to the rule of law.

It isn't like the FDA can just NOTICE that they are a net negative on the medical industry and simply CHOOSE to stop the process of limiting medical innovation without doing cargo-culted science first: they are literally legally required to do this.

If you absolutely insist on blaming a person... maybe Kefauver (who died in 1963)?

What these bureaucrats could do is not fight reform of their totally broken bureaucracy, in the press, when their legal mandate is noticed to be insane and subtracted from the law codes by thoughtful legislators elected by a minimally tolerant median voter.

However, in the case of the FDA specifically, they did the opposite of this, and just had enemies lists. 

The Times' report also details evidence of an "enemies" list within FDA-"unusual at domestic agencies that do not handle classified information," it noted. The documents detailed efforts to fire employees within the agency that had gone public with complaints, attempts to track contacts with members of the media, get in front of potentially damaging stories by assessing which members of its staff were contacting media outlets and rank members of congress deemed to be unfriendly towards the agency.

Because of course that's how it worked! Why would we imagine anything else in a system as broken as ours, that has incidentally killed as many people as ours has?

This is actively treacherously-turned mesa-optimization, implemented in human bureaucrats, with respect to their proper governors: the elected officials and the voters. 

This is anti-democratic regulatory capture on top of bad legislation. 

This is a bad institutional "adaptation" being staffed by bad people, where anyone good and smart would resign because the law commands objectively stupid behavior, and dumb people might just turn the crank anyway, and anyone really smart and bad could climb to the top, use the power for profits, chase elected officials away from spending political capital trying to reform them, and eventually leave for a cushy lobbyist job. This is organizational cancer. This is the FDA.  Each part of it is rotten. Every new closet you look in... you find a new skeleton.

Maybe the other bureaucracies are good? On base rates surely some of them were set up well, and still serve a sane purpose faithfully? It would be kind of remarkable if ALL the bureaucracies were so dramatically "the opposite of good". You'd have to look at each one individually I think. I haven't looked at all of them?

I've ready good things about the founder and founding of the naval nuclear power program, if someone wanted to do a deep dive on a potentially positive case.

Of the ones I've looked at, the FDA seems to me like it has the highest combined score of roughly: (("really really  bad") x ("obviously so, to very many people, because of a Recent Event") x ("deeply instructive such that getting the right answer about the FDA from first principles probably implies LOTS of other probably useful models for thinking about many similar issues")).

Here are a couple of very simple points that I don't think should be controversial which I think are useful to frame the discussion.

First, there is a linguistic issue: what is a bureaucracy? Is it any large organization? Is it a rigid organization? Is it a corrupt organization? People use it in all these ways.

Blame-minimization is a theory of bureaucratic decay. If all bureaucracies are bad, why do they exist? If everyone agreed that they are bad, why don't we tear them down? They are sometimes created to provide fake jobs, but usually they are created because they are expected to work and the attention at the beginning makes them work. You can't explain them with just a principle of decay because that inherently can't explain the starting point, only the trajectory. Compare Samo Burja.

Interesting. I worked in and with a few bureaucracies in NZ and I very much doubt there is a single model to explain or predict behavior, because multiple utilities and motivations are present. They are plagued (as are private companies) by the levels problem where information between levels of management can get twisted by differing motivations and skill level. As other commentators have pointed out, upper levels of the management can be extremely risk adverse because they crucified for mistakes and unrewarded for success. While "blame-minimization" might seem appropriate, there are other factors at play. Large among them would be motivation. Some bureaucrats are empire-builders and their utility function is ever-increasing areas of control, (career administrators in middle-management role) but others got into the game in the first place because they wanted to change the world, and the tools of government seemed like a good place to find power. With that kind of motivation, they tend to rise quickly and I see a fair no. of them in high positions, especially in education, health, welfare. They feel the forces of blame, but are individually motivated to make change. Good luck predicting outcomes there. 
The other prediction problem would relate to where in the organization that a decision is made. The more technical the decision, the more likely that is being made at low level in organization among the technocrats. The decision may still have to percolate up the levels which it may be misunderstood or subtly reframed to make a middle manager look good, (another predictability problem) but mostly I would expect such decisions to reflect perceived technical utility. (eg best timing for a booster vaccination). 

A bureaucracy works well when every person has a vested interest in the shared success more than in whatever Goodhart incentives tend to emerge in the bureaucratic process. An essential (but by no means sufficient) part of it is the right amount of slack. With too little slack the Goodhart optimization pressures defeat all other incentives. 

Great post!

I think it's good to first expand a view on bureaucracy a bit -- while FDA may look as a prototypical bureaucracy, it's not the only example

As Matthew Barnett said, for-profit firms also create internal bureaucracies. In Soviet Union and other communist countries all command economy could be viewed as a single firm, and all economic management was done via in-firm internal bureaucracy. 

So I'd hypothesize that the thing bureaucracy (at any level or subunit) tries to maximize is "power", or to put it more concretely, ability to give orders. In that case an attempt to maximize its budget, or extend its purview evidently gives a manager inside a bureaucracy more capabilities to give orders, both inside and outside of bureaucratic hierarchy

In this framing avoiding blame is more like instrumental and defensive goal -- being constantly blamed for some transgressions may cause higher levels on hierarchy (or principals/politicians/firm executives) to reduce manager's or unit's power.

For whatever it is worth (probably not much) an asosciation what the thing could: acceptance-maximization.

Some intuitions in this direction: Inducing lethargy seems somewhat unique to this class of optimisers. Why blame seems to be dangerous is that it leads to rejection of the status quo. It seems it is common to have a attitude that the bureaucracy touch is both painful and neccesary. This seems like a hard sell and the other optimisers do less of it.

Very interesting post.

This may or may not be relevant, but your post reminded me of Venkatash Rao’s Gervais Principle, where he describes a hierarchy of sociopaths, clueless, and losers. It’s based on corporate hierarchy, but maybe it might also provide some illumination on bureaucracies? 
https://www.ribbonfarm.com/2009/10/07/the-gervais-principle-or-the-office-according-to-the-office/

A bureaucracy works well when every person has a vested interest in the shared success more than in whatever Goodhart incentives tend to emerge in the bureaucratic process. An essential (but by no means sufficient) part of it is the right amount of slack. With too little slack the Goodhart optimization pressures defeat all other incentives.

Bureaucracies are sheltered institutions, that is institutions whose existence is not threatened by their own poor performance.

Institutions are composed of overlapping [goals / mechanisms / cycles of behavior]. The most universal of these goals is self-perpetuation or continued existence.

Each clear example (firm = profit motivated, politician = vote motivated, animal = food motivated) ties back to a respective input for continued existence. Almost universally, existing entities strive against entropy to continue existing, which makes the inputs required for their existence a useful and widely applicable model of their behavior.

Bureaucracies are sheltered. Their existence continues as long as they clear artificially low expectations. They are often protected by larger more functional entities, such as an internal bureaucracy inside an overall productive company, or a house-cat being fed by a human family.

Some bureaucracies are functional. Especially early in their existence, the overlapping [goals / mechanisms / cycles of behavior] in a bureaucracy will be relevant to the problems and technologies on hand. House-cats are still evolutionarily good at catching mice. But as the environment changes, bureaucracies face insufficient pressure to adapt, and thus tend towards dysfunction.

There is no broad "bureaucracies are X-motivated" statement because bureaucracies have no mandatory self-continuation loop, so they lack a common motivator.

f you go look at real bureaucracies, it is not really the case that "at each level, the bosses tell the subordinates what to do and they just have to do it". At every bureaucracy I've worked in/around, lower-level decision-makers had many de facto degrees of freedom. You can think of this as a generalization of one of the central problems of jurisprudence: in practice, human "bosses" (or legislatures, in the jurisprudence case) are not able to give instructions that unambiguously specify what to do in all the crazy situations which come up in practice. Nor do people at the top have anywhere near the bandwidth needed to decide every ambiguous case themselves; there is far too much ambiguity in the world. So, in practice, lower-level people (i.e. judges at various levels) necessarily make many many judgment calls in the course of their work.

The fact that blame-minimalization seemed somewhat plausible as a first idea already tells us that the question might be unanswerable:

All these variables to be maximized or minimized are relevant for the respective theories in which they are central because they determine some sort of quality or at least some sort of measurement that enables the respective entity to either be better or worse than actualy alternatives (competition between companies or politicians) or at least it may serve to enable others to hold the entity responsible for being better or worse than actual or imaginable alternatives (and therefore maybe be replaced by another one etc. etc.).

This automatically leads to a selection-mechanism that shapes the respective entities - though the similarity to natural selection is at least in some cases pretty limited.

Since a theory of such entities is suposed describe them and therefore the shapes they have/take, it must at least in some way deal with (facts that depend on) the way those entities are being shaped.

But if there it is really at least to some degree true that bureaucracies tend to limit blame(worthyness), then this entails that they tend to limit their own responsibility (in some regards). What follows is that - at least to some degree - they lack the respective shaping-mechanism, the selective pressure that follows from sometimes being blamed and therefore sometimes being replaced by an alternative. 

What follows (and seems very plausible at least to me) is that the dominant shaping-mechanism of a bureaucracy is just the sum of decisions that lead to its installment (again, this is only the case to the degree that this lack of blame-worthyness is given). 

Such decisions (correct me in so far as I am wrong, of course) are based on specific laws and regulations that ought to be implemented.

Therefore, when we try to formulate a theory of bureaucracies, we deal with a whole lot of individual differences when it comes to the defining characteristics of the entities that ought to be described - due to the differences in terms of those laws and regulations.

I therefore suspect it will at least stay very hard to formulate an adequate theory of bureaucracy - at least if we stay in a realm of such simplicity where the maximization or minimzation of a single variable is our starting point. 

Of course we might say "the degree to which rule x is adequately implemented is being maximized" might be a fitting description, but the blame-avoidance makes it impossible to say that this rule actually describes the entities we want ot describe and not just the intentions that went into their respective installation. 

(sorry if my writing is kind of inefficient - I fear this might be the case - as I am not a native speaker)

If you really want to learn about this in a ton of detail and spend hours and hours reading about it, look at Dominic Cummings’s Substack and previous essays. Also his interview before parliament is good on the UK’s handling of covid: https://youtu.be/8LFS3FaRs_s

He was Boris Johnson’s chief advisor and actually espouses a lot rationalist/EA ideals, he reads a lot of rationalist aligned stuff etc, so it’s quite instructive to have a view into British bureaucracy through him.

From what I’ve read the main impression I get is that the quantity bureaucracy maximises for is… status quo. Keeping things however they are no matter the consequences is clearly the goal of much of the state, even at great human, material, monetary, and time cost. This is why there are no long term goals in government, why there are constant catastrophic failures with no change, and why getting the right people in the right positions is nigh on impossible.

I can’t find it atm but there’s a quote from Cummings when he worked in the Department of Education and was trying to bring about change, where a senior official said to him “you’re a mutant virus, and I’m the immune system trying to keep you out” or something along those lines.

Bureaucracies resist change at all levels and at all costs.

There was a time that the distinguishing characteristic of working in a bureaucracy (as an employee) were: it's not very fun; the pay is OK but not great; the hours are good (no nights and weekends), the benefits are pretty good; it's difficult but not impossible to get fired; and the retirement plan is very good (both benefits and pay). In that world, I'd argue that the vast majority of employees in any bureaucracy -- all but the top handful of political appointees -- focus on optimizing their own longevity as employees. That can explain a wide variety of behaviors -- while it probably leans toward CYA behaviors at steady-state, there will be times that an employee would get a signal from above on a non-CYA maneuver that would be a good thing to do. So the surprising nature of the actions taken could just be a reflection of a personal preferences, political events, or even a mood from the top (very small group of people) that can change fast because that group is small, non-expert, and political. And that preference is magnified by the 99.9% of the bureaucracy's employees who are watchers and followers. No long-term policy preference or coherent strategy other than, let's do what the top people seem to want right now.

The fact that more recently bureaucrats are actually paid well suggests the above would be even stronger tendencies now. Further amplified by social media -- you don't have to just satisfy your bosses; you also have to satisfy the vocal people your bosses seem inclined to satisfy. The last element is why care if you're hard to fire? I think the answer is, you still have to put in a lot of years to get that prize retirement. You could either do that in relative harmony with your co-workers and sometimes get assignments that aren't a grind; or you could be the person other people don't like and have to work through many years of social exclusion and bad assignments.

Bottom line: keep your job forever; and not have it be more bad than it has to be; by implementing the preferences of your bosses in a way that's visible to them and others -- whether it's effective or not.

Someone should make a game/simulation of these things. Let the layman learn how to navigate politics and let the sociologists plan better.  This is the only way to get real answers, given the extreme political nature of the issue. Ok, so there is https://en.wikipedia.org/wiki/Social_simulation but the games for laymen (like the sims) listed are mostly bad and certainly do not simulate office politics. 

Bureaucracy exists in the public sector because the government has certain duties to the public which must be fulfilled, and the process of fulfilling them is inherently complex, difficult, and expensive.

It is also quite labor-intensive. Many people are required to facilitate the government's administrative efforts. From court clerks to prison guards to the person who snaps your driver's license photo, public service requires a lot of grunt work, and the direct impact of their performance upon everyone's individual lives necessitates constant meta level analysis. This is further complicated by conflicting interests, disagreements over priorities, the prevalence of logistical errors, funding and informational deficiencies, social fads, internal ambivalence about the institutional mission, efforts to eliminate corruption and waste, corruption and waste period, and many other variables. Managing all of this at the federal, state, and local levels is mind-bogglingly convoluted and inconvenient, and I think necessarily bureaucratic.

In other words, the enormous difficulty of sustaining a functional society more or less justifies the existence of bureaucracy. The point of having a society is to increase the collective fitness of its members in the interest of improving their overall quality of life. Of course, sustaining a somewhat coordinated social framework is not a perfect solution to the problem of trying to be alive correctly, but it's probably better than not doing it.

So you might say that bureaucracy in the public sector is, broadly speaking, an "efficiency-maximizer" with regard to public administration (at least in theory, although in practice bureaucracy is widely associated with inefficiency, which indicates that it's simply failing to achieve its goal); or, even more broadly speaking, a "fitness-maximizer" that doesn't happen to work very well, yet persists in the absence of anything better that could reasonably be expected to overcome entrenched resistance to reform efforts.

Whatever variable bureaucracy exists to maximize, its failures to do so are obviously not a reflection of its goals. All corporations are profit-maximizers, including those that eventually file bankruptcy.

Everything would be easy if only it weren't so damn hard, etc.

The blame minimizer reminded me of the situation of the medical profession in the US and the UK, where the individual action is strongly shaped by the risks of entering costly litigations. The Hipocrates oath (do no harm) does not seem to be enough to steer action towards a maximum utility for the patients.

To remediate the status quo and reduce the risk-averse mentality, maybe we need some kind of protection from undue pressures (lobby, idealogy, short-sighted politics)? In a few Countries like Germany and Brazil, the public officers cannot be easily fired and are defended in court by competent public office lawyers. This is not the case in the US, where even judges, prosecutors, and sheriffs are elected and, hence, have to please their voters/sponsors between election cycles. 

There is no structure of incentives that would solve the problem of blame-minimization if the individuals making even the small decisions are afraid of doing the right thing. Furthermore, it is really difficult to retain the most honest and the brightest in public office when the environment is driven by power-grabbing (current Republican politics), where doing something for the greater good is contingent upon a personal risk/benefit calculation.