A Taxonomy Of Bureaucracies & Science

I tend to say tongue-in-cheek that science needs more bureaucracy. But I realized that I can develop a 2x2 matrix to classify different ways in which bureaucracies organize to communicate what I mean. Let's say bureaucracies can be classified to be autonomous vs procedural, and flat vs hierarchical.

There’s actually a 3rd dimension: static vs dynamic. I’m wary of going beyond a 2x2 matrix but I think all 3 dimensions are necessary. The matrix describes the organizational structure and the latter the adaptive capacity. Bureaucracy is desirable because that’s the only way to achieve things at scale. We too often use it as a pejorative which perhaps blinds us to the need to build more functional bureaucracies. Though I think this is not an alien concept to introduce to this forum.

In my categorization, NIH and NSF funding for example can be thought to be flat, procedural and static. DARPA (at its best) autonomous, hierarchical, and dynamic. Of course DARPA is hierarchical relative to NIH, not necessarily in an absolute sense. There are program managers in charge of a budget who are responsible for outcomes and who are empowered to make decisions. They don’t defer to peer-review. Perhaps bell labs and Deepmind pre-LLMs were flat and autonomous.

This taxonomy applies to militaries as well. The Soviets were hierarchical and procedural. Germans hierarchical and autonomous (mission command) Guerillas tend to be flat and autonomous. And perhaps UN peacekeepers are flat and procedural.

The reason I say I think science needs bureaucracy is because I think there needs to be people responsible for making decisions on what sort of infrastructure to build: scientific tools and software, which requires coordinated engineering effort, not novel research. Likewise, we need a QA step to check for replicability and to hold people accountable if their research is dishonest. And need to establish better publication norms e.g. pre-registration of hypothesis.

Whether a bureaucracy is static vs dynamic depends on the concentration of live players in critical positions in a bureaucracy. And that perhaps is downstream of whether this organization is subject to selection effects. By that I mean there is a competitive dynamic / cybernetic feedback amid changing circumstances.

The US military for example is a very competent organization, hierarchical and autonomous. But its lacking in dynamism (perhaps because they’re not in charge of their own procurement).

These different bureaucracies can have characteristic failure and success modes: flat procedural organizations fail through entropy, hierarchical procedural organizations through ossification. Flat and autonomous organizations through a failure to coordinate. And hierarchical and autonomous organizations through bad judgement at scale.

To circle back to my initial claim then, what I mean by more bureaucratic is more hierarchical (but also autonomous and dynamic of course).

Good post. I would urge dropping the tongue-in-check "more bureaucracy," as it can more readily (and reasonably) be interpreted to mean something clearly wrong, than the interesting, potentially correct meaning you go on to explain.

This is an example of the temptation to use cute, oppositional wordings; the temptation should be avoided. The payoff is rarely worth the invited confusion.

Eh. Every time I read something by an actual researcher, they complain about too much centralization. Also, generally the centralized Big Science projects of the last few decades haven't been that successful. The Large Hadron Collider failed to discover any new physics (although I guess it helped confirm that the Higgs Boson existed). The Human Genome project didn't lead to anything much in terms of practical applications. Anyone remember Obama's BRAIN initiative?

Contrast that to something like AI where there are dozens of labs all working on their own models.

Even the Soviets knew that competition was good when applied to the military sphere. They had multiple design bureaus for tanks and aerospace systems. Often this resulted in better designs.

Imagine if there were multiple US rocketry bureaus instead of NASA. One of them might have come up with something better than the Space Shuttle. Of course, now that we have private space companies, they've done exactly that. But we could have done it even in the public sector if we were just willing to embrace less centralization.

I agree that we need more people working on replicating scientific results. It's dumb that we have so much reverence for professors (pretending) to teach bored undergrads, and so little reverence for finding out if the science we're doing is actually, you know, true, by replicating it.

>>5314
I suppose I do prefer my own taxonomy to 'too much centralization'. One can distinguish between different organizations on how hierarchical and how procedural they are as independent axes. Granted neither are a clean binary, you want a diversity of ecosystems, and different fields and institutions should have a diversity of configurations but you don't want to be maximally one way or the other. That said I think on the margin scientific research organizations should be more hierarchical and less procedural. In fact the AI labs are often organized this way.

With respect to NASA, well look at how its being reformed: we put Jared Isaacman in charge, and told him to turn it around. There were already supposedly competing NASA centers. The problem is they weren't actually competing because the government funds them to be jobs programs and they were shielded from cybernetic feedback.

As to the science projects, the BRAIN project didn't set up a new institute or a bureaucracy it was a 'cross functional' and 'collaborative' effort to write grants the same old way subject to same old peer review structures. How can you cite that as an example of hierarchical organization. Ditto for the cancer moonshot or ITAR or other wastes of money. Compare the org chart of those initiatives to the Apollo and Manhattan projects.

The way science is structured is already extremely hierarchical, though. Go to any biology or physics program. You'll find a small number of PIs (principal investigators) and a large number of graduate students. The grad students are basically doing what the PIs (and maybe to a lesser extent the postdocs) tell them to do.

I mean, the grad students get some flexibility in what they research and how they do it, but at the end of the day, if their advisor doesn't agree, they'll have to find a new advisor. There's really not that many labs at each school. It's not like everyone is doing their own thing and going off in a totally different direction.

The problems I see in academia are apparently very different than what you see. One big problem I see is the grievance studies departments that have grown like tumors in the last few decades. There are whole fields which are just total nonsense, people exchanging nonsense word salads that could be generated even by the dumbest 2020-era LLM.

In fact, I would go so far as to say that "the humanities" (I hate that term, btw) should just be totally shut down in their current form. We can bring some of it back later, but NOT with any of the current clowns involved.

Science has a different problem, which is the tendency to rathole on a small number of unproductive topics. For example, how much ink has been spilled over string theory, and where did it get us? If anything, we should we should give scientists more leeway to explore new ideas, as long as they can be rigorously connected back to the physical world (that last part is very important).

Engineering is one area where megaprojects and centralization makes sense. But in order to get to the stage where that would even make sense, we have to solve a lot of the other problems. For example, building new designs for nuclear reactors, large-scale desalination, or nuclear-powered rocket propulsion, are excellent projects to fund. But how can we get there without fixing the current political situation? The engineers are doing the best they can with our current whiny loser political class.