Bureaucracy as Active Ingredient
slatestarcodex.com
slatestarcodex.com
So then he raises the archetypal story of the DMV. For being the primary identity providers of the US populace, the DMVs of US states are actually rather remarkable. I guess the seemingly arbitrary progression of ticket numbers in cavernous waiting rooms is valid annoyance, but watermarked cards being mailed out to thousands of people just a few business days after collecting their info is pretty good, and then there's more vehicles than that.
Bureaucracy gets a bad rap, but it's worth applying the "I could've built that" test for them the same way it's trendy to critique valuable companies that are "just a chat app". If you had to credential 35 million people and 39 million vehicles a year, how would you do it? Would you codify a few things, and hire rockstars who can think on their feet? How soon would you start making rules about the blatantness of invalid forms, or unpaid fees, or how many people would you have to catch selling favors before you re-examine the process?
Deep bureaucracies are often the emergent result when your process has been burned by costly customers (in time, money, effort) or rogue employees (corruption, graft, espionage) too many times. They are systems where each and every decision maker can always defer to the Process as an explanation for why someone can't have something and the Process is sacred because "it's always been that way", even if it changes, and where each and every fulfiller is powerless to truly make a meaningful exception, except when that feature was already built in.
> So then he raises the archetypal story of the DMV
Maybe read it again? The author raises a coherent hypothesis, then questions it as a widely explanatory principle, using the DMV explicitly and intentionally as an example of bureaucratic inefficiency that it describes poorly.
We should remember sometimes that not everyone thinks in terms of bandwidth and scale.
Time and exasperation people are willing to put up with in order to get what they want is a form of co-payment. Part of it is paid for by the insured (as a "proof of real need"), another part being the bureaucracy which is paid by all insured people. Any insured may prefer to wait (than to occupy the bureaucracy by filling useless forms...) because it is far cheaper (no work for him nor for the bureaucracy), but this is rarely possible because there are urgent cases, and also because it seems more frustrating.
Unbelievably, the medication example the author uses is a system I have direct experience with. It is a group whose job it is to make the decision of whether insurance covers a drug that a doctor has prescribed.
The triage is more sophisticated than a queue management ritual, but also has elements of it in it.
Some insurer programs require that the patient be declined by another one before paying for the prescription, so the doctors fill out hopeless requests because they need the rejection letter to achieve their outcome. This fills the decision queues, and everyone experiences the lag. Consider doctors handwriting as a next order CV problem, and then think about the attitude of the user demographic who doesn't feel obligated to be clear about their intentions - now design a digital product for them.
Bureaucracy in real life is a kind of cluster-governance, where nobody serves each other, and everyone presumes everyone else is coming to them to be regulated. Further, they really do think their job is to avoid obligations to others unless they are somehow mandated to have one.
I can say that agile development methods are disrupting bureaucracy in a big way. You don't see it, but every online license renewal and filing system wipes out layers of bureaucracy. The systems take years to develop because the personal/business relationships that made those systems work need to be identified and unraveled before you implement them in code. I would argue you don't understand stakeholder management unless you have spent time in a bureaucracy, and the people who do succeed at it in the private sector often do so as the result of both luck and negative anti-patterns.
I saw how the internet changed government almost 20 years ago, and can say that cloud and microservices are now also revolutionary.
Tech is an expression of how its users want to relate, and in this way it's culture. When we impose tech on institutions, it realigns them to organize around the relational patterns in the tech.
Bureaucracy is an active ingredient, and the tech we develop changes it. Want a better DMV? Build one.
Please elaborate. I would be very interested to hear more about this from your perspective. Thanks!
It created the abstractions of clients and services and put enterprise architecture in the middle. All projects became IT projects.
Today, the clash of iterative methods that provide makers with almost total autonomy vs. waterfall PM oversight and governance is the key friction point in large institutions, imho.
I work in a very compliance heavy environment, and we run an “agile” team that ships as often as 10 times a day. We have easy audits and no compliance issues because we invested in good process design. A company we’re quite close with has the same compliance requirements, and their shipping process involves printing, signing, scanning and filing excel sheets. It’s not like we’re vastly better at our jobs than they are, we just designed far better processes.
Company bureaucracies skip this part as matter of course. Often the adherence parts or existence of workarounds.
"Orgs have an incentive not to respond yes to every request they get" - not true. Orgs have an incentive to reduce the cost of saying yes, but saying yes is an important part of providing good customer service, keeping repeat business, and building a brand that enables bigger growth. Orgs that lose sight of this, deserve to be eaten by their competitors.
It's also wrong to assume that just because a company gets acquired, that there's going to be 'unnecessary levels of administration and other associated costs' that are just begging for the ax. Pharma companies have finance people too that can provide that kind of oversight and recommendations.
The market for drugs is complicated by two factors. First is that people who need drugs generally can't get that need met by any other product / service. So there's an incentive for price creep. Second is the patent system which forces companies to grab as much cash as they can until the patent expires.
If longer patents were granted, or the patent system were more sane in that the patent period doesn't begin until after the drug is able to be made available to the market, then the costs of developing drugs would be a lot more legible, and therefore amortizable.
Also are you seriously arguing here for longer patent terms? You think that will get new products to market more quickly?
The irony is that bitcoin, the pioneer, confirms your hypothesis because it relies on a form of intentional waste (POW) to disincentivise bad behaviour. PoW is the new bureaucracy.
I do think we will find better ways, and decentralisation is the natural outcome.
On this topic I would recommend one of Scott's most well-known essays, Meditations on Moloch [1]. The meat of it is this:
> …in some competition optimizing for X, the opportunity arises to throw some other value under the bus for improved X. Those who take it prosper. Those who don’t take it die out. Eventually, everyone’s relative status is about the same as before, but everyone’s absolute status is worse than before. The process continues until all other values that can be traded off have been – in other words, until human ingenuity cannot possibly figure out a way to make things any worse.
If top-town authoritarianism is at one extreme of the spectrum, then completely decentralized systems is the other undesirable extreme, equally bad if not worse: if there is no regulation, then there is no strong coordination, and a race-to-the-bottom situation ensues (in optimizing for X). If everyone is acting game-theoretically optimally, then anyone who tries to improve the situation gets eliminated. It might be locally optimal, but it's highly unlikely to be globally optimal.
Decentralization means being a slave to no one, which means being a slave to the incentive system, which means being a slave to everyone else.
But I would also encourage thinking of it not as a dualism, a strict choice between bureaucracy and blockchains, but as a spectrum instead. This means that there's an entire world of valid choices in between. I would very much argue that the optimum can be found just by moving closer to the middle.
[1] http://slatestarcodex.com/2014/07/30/meditations-on-moloch/
My thought here is that improving situation doesn't have to happen in the game. One can step out of the game and offer a new one.
I.e. the quoted judgement about decentralised systems is not as grave if you don't look at them as closed systems.
This isn't true at all. The fact that the efficient market hypothesis doesn't seem to hold is already the proof of that, and the same reasoning applies to other social behaviors [4].
Furthermore, many, many, many, many studies in neuroeconomics and behavioral economics (and related fields) [1] have shed light on the bounded rationality of human actors in decision making. Game-theoretically optimal decisions is a specific, well-defined thing that's different from making decisions based on the affects and biases of the preconscious and unconscious. You can't just tautologically define "game-theoretically optimal" to be whatever decision someone ends up making.
What does this mean? The direct implication is that it's neither game-theoretically optimal nor inevitable for stateless humans to enter into a social contract that eventually cements itself into a state apparatus.
A top-down authoritarian doesn't get their powers from a deity, yes, but it wasn't quite explicit coordination either. To demonstrate using a brief example, one of the models of the evolutionary stages of the social contract [2][3] is that states were a natural conclusion of tribal societies, which were themselves a natural conclusion of lineal societies, i.e. families. Lineal societies most likely weren't formed through explicit consent, i.e. early stateless infants probably didn't explicitly decide to enter into a social contract. There's basically no particular a priori reason why families exist (for example, the Ik people in Uganda seem to have no or a very weak concept of using familial relations for survival, even to protect their young), and very likely that this evolution was just the model happened to win out due to random noise in history rather than actual optimality.
Again I would like to stress how much the current societal configuration is mostly just noise instead of signal.
[1] For an introduction, https://www.goodreads.com/book/show/11468377-thinking-fast-a...
[2] https://mitpress.mit.edu/books/nomadology
[3] https://www.goodreads.com/book/show/118317.Anti_Oedipus
[4] http://slatestarcodex.com/2017/11/30/book-review-inadequate-...
Yet now you write that everyone acting game-theoretically optimally is not actually possible.
So you were criticizing an impossible scenario?
Since I agree with you that literally everyone acting game-theoretically optimally is impossible, I took your earlier use of the concept to mean something possible, that is, everyone acting as close to game-theoretically optimally as they are capable, and used the concept in the same way.
The multipolar trap manifests itself in real life in lots of simple ways already. The original essay has some examples, but the tragedy of the commons is the quintessential example.
The multipolar trap as a totalizing behavior is not an impossible scenario, but just currently infeasible for most sufficiently complex scenarios given the bounded reality. Which means it becomes a cause for concern as the bounds on rationality increase (through computation, through organization, etc.). That's why we should be wary of decentralized systems in the long run.
What you said was "everyone" was always acting game-theoretically already and therefore that what we currently have is game-theoretically optimal, which is what I disagreed with.
Coordination doesn't imply regulation, conventions (which doesn't imply any chief/judge) are sufficient.
Top-town authoritarianism is IMHO worse than decentralization.
Some of them would call or email me with some really important, not previously discussed thing that they absolutely needed to have (but really didn't). Where in the past I'd immediately take this request seriously and discuss how many hours/days of work this would reasonably take, I now often give them an unusually high indication at first. If the conversation continues and it becomes clear that they really do want/need said feature, I'll come up with a more realistic indication.
1) thing is not essential, you saved the cost
2) a workaround that is inconvenient or costly exists, compare costs
3) you just caused an opportunity cost
In other words, you're breaking the process. Unless you're a process engineer and track both causes and results diligently you should not do that. Ever.
Obviously one should use common sense to avoid adverse outcomes. Inexpensive productivity tools get approved more quickly, whereas upgrades to perfectly fine but previous generation laptops may take a week or so :-).
Then the role of some person should be to gather such tacit knowledge and convert it into a usable and transmittable form.
Now if they do not know about one, they probably should be informed. :)
Perhaps you're not filtering out requests on the basis of need, but on the ability of some individuals over others to figure out the unwritten rules and choose to play your game.
Yes, it is possible that people are gaming my tactic and I am optimizing for the wrong variable, but in practice that doesn't seem to happen.
http://dilbert.com/strip/1995-07-28
http://dilbert.com/strip/1995-07-29
The "approved list" as a philosophy or state of mind remains one of the best approaches.
Certainly, once a bureaucracy got to that point, it might have certain benefits, but I don't see how that implies there was a conscious strategy before.