11 mental models of bureaucracies

What happens when tens of thousands of people try to coordinate towards an ever changing set of competing objectives?

What is going on?

My work focuses on helping people figure out how they can have an impact in the Civil Service. Questions that I, and the people I coach, ask a lot, include: “What the hell is actually going on here?” “How are decisions being made and why?” “Who has power over what and what are the dynamics of that?” “What games are these people playing?” The system we are working in seems to work as it’s supposed to some of the time, but not others. It seems inadequate in a bunch of ways, but it’s not obvious what the root cause is, or what that means about how to navigate the system, or make it better.

I think this stuff is very difficult. There is clearly lots of politics (both institutional politics and politics politics). No-one is really incentivised to tell you what they think is actually going on unless they need you and trust you. People also don’t tend to tell you whether they trust you. And trust seems to be generally low in big, complex organisations.

I am uncertain about all of this….

To try and understand how the Civil Service works, I took a step back and looked for models of how bureaucracies work more broadly. I then drafted the below list. They vary a lot in scope, and how much of the bureaucratic dynamics they explain. There isn’t a clear taxonomy of them, and they don't fit into a neat single theory. I think they all add something, but I am more convinced of some more than others.

I worry a little about some of this coming across as critical or cynical. But I’d emphasise that I don’t think the existence of these dynamics should be seen as a criticism of any individual in the system. People don’t create their incentive systems. The powerful are often less powerful than they seem, and they also deserve our compassion. It’s also important that we are able to look clearly at the anti-social dynamics of systems if we ever intend to make them better

11 mental models of bureaucracies

Stakeholder paralysis - the idea here is that the greater the number and complexity of the stakeholders that you have, the greater the difficulty in making decisions and implementing changes. This is due to having a wider set of conflicting goals, and because stakeholders have the power to block changes that are not in their interests. The implication is that to get things done you need to develop a really clear picture of your stakeholders and possibly try to minimise them in number. This is also one of the key points in The Dictator's Handbook. I felt this viscerally in my roles in the Cabinet Office. As I was drafting some cross-government standards, it felt like I had my >200 person stakeholder network watching me over my shoulder. I had an intuition for their needs and the things that they would complain about, and those intuitions controlled me. A related point here is the importance of looking for and managing “veto players” - stakeholders who have a lot of power to stop you doing what you’re doing. Your management chain and minister are usually all veto players, but other stakeholders will be too.

Social construction of reality - this describes how the information landscape is manipulated to benefit those with power over the information landscape. All groups have competing sets of internal ideologies (beliefs about how the system you’re in works). Powerful people in the group are incentivised (probably mostly unconsciously), to prefer those ideologies that reinforce their power. A classic example of this is a senior manager emphasising the idea that the system they manage is meritocratic. The ideas that those in authority deserve to be there legitimises and strengthens the manager’s authority. This sounds kind of awful, but close up this can feels much more vanilla. This might just look like each individual taking opportunities to talk about their successes while downplaying their failures, which everyone is incentivised to do…

Moral Mazes - suggest, among other things, that a primary goal of all individuals in a hierarchy is to “make your manager look good”. Not least because your potential future managers will be watching you for signs of disloyalty to managers. It also talks about how a lot of communication in an organisation is actually euphemistic double-speak; individuals are incentivised to push the landscape of power and influence towards themselves and people they trust, but need to maintain plausible deniability to do this. I.e. they need to abide by all the other social norms, which include not game-playing, not lying and not stabbing people in the back. An example of this might be describing someone you think would make a bad manager as “very detail oriented” as a way of implying someone is bad at seeing the strategic big picture and therefore not suited for senior management. And a classic example of this in the Civil Service, is basically anything being described as “interesting”, to mean “bad”, “strange”, “awful”, or “I hate it!”. One caveat to this is that in a more senior and more competitive environment, the plausible deniability becomes harder to maintain (because you end up with more common knowledge of these tactics). 

Blame minimisation and prestige maximisation - Individuals are incentivised to maximise the expected level of their prestige and minimise the chance of any embarrassing failure. This results in individuals and teams manoeuvring to get impressive remits and projects that have low risk of failure or where failure can be blamed on something or someone else. Risk aversion is a common complaint about the Civil Service. This model seems particularly useful when pointed at politicians, where it might be close-enough to just think of them as expected-vote maximisers. This likely affects almost everyone to some degree.

Seeing like a state - this is the idea that governments tend to push systems that they govern towards greater legibility. So that the system can be better controlled (and taxed). Where a system is illegible, and the state tries to intervene, things can go very wrong. This probably generalises to other lower-level power structures - e.g. a senior manager pushing their team towards doing work that the manager understands or can measure, at the cost of productivity (i.e. they push their team towards “Goodharting” - where the team becomes incentivised to just do the thing that is being measured). 

Principal–agent problems - this is key system failure that results from decision makers being different people or groups to those their decisions affect. A lot of the systems we use in government are attempts to minimise these problems (i.e. increase accountability), but it is pretty fundamental and affects everyone. Bureaucratic/regulatory capture is a broad example of this. A specific, sad, example might be a senior leader who is in a position to be able to steer the conclusion of the monitoring and evaluation of a program that they are managing.

Inertia - Risk aversion, a lack of any individual or group having a clear overview, and high costs to changing highly integrated systems all incentivise interia. This means the default is for the system to keep doing what it has always been doing.  The cost of changing your plan during a project is really high. If you have a plan that is heading towards “something that looks vaguely like not-failing”, it’s going to feel very risky to switch approach. I worked on one large tech project in government and this felt real. No-one really wants to hear about anything that sounds like “we need to change what we are doing”. We are there to deliver something; not something great. 

Coordination headwinds - the basic idea here is that groups get more complicated and uncertainty about the chance of project success grows, it becomes harder and harder to convince people to complete work. The underlying concepts of, volatility, uncertainty, complexity and ambiguity (VUCA) seem to be key variables in whether a system is going to be able to actually do things. The Cynefin framework is an attempt to address this.

Live vs. dead players - suggest that a key variable to measure teams or individuals in an org is “agenticity”. Or how much they are willing and capable of doing new things, really understand how things work, and expand their influence. Most teams and individuals are “dead” and not really capable of doing new things. This makes them predictable. Live players are hard to spot, because they generally try to hide in order to avoid competition. But they can be identified by the fact they are attracting impressive people and moving into new areas. I like this model, but suspect it’s a little “Myers-Briggs-y” - (i.e. creating compelling-seeming categories where there is little value in doing so). The lesson for me is that people and teams probably vary massively in how much agency they demonstrate. It’s wise to try and assess agency to figure out who you need to coordinate with and watch out for. 

The Gervais Principle - This provides a (slightly unnecessarily) cynical take on the relationship between senior managers, middle manager and junior staff. Venkatesh describes:

• “Loser” junior staff being largely concerned with social competition.

• “Clueless” middle managers who obsessed with doing well according to the legible rules of the system

• “Sociopathic” senior managers who see through these rules and social games in order to dominate others

I enjoyed this series, and I thought it was worth reading (would maybe recommend starting with Scott Alexander’s summary/critique though). I think it probably goes too far in a lot of places, and is also a little unnecessarily Myers-Briggs-y. 

Punctuated equilibrium - This one feels a bit specific to politics, but I think it can explain a lot about how to get things done in a complex bureaucracy. Paul Cairney talks about how nothing happens most of the time, and then suddenly the policy area gets to the top of the agenda and options are needed urgently. Because of this pattern, it's not wise to delay developing policy options until there is political interest. My experience of this is that I felt like I regularly had good ideas for change in government, but my attempts to push these from below or convince senior people failed. The times when I was successful were when I had separately been working on an idea and then for some external reason suddenly there was a reason senior people needed a solution to the problem. The lesson was to work on solutions before you need them and be ready to deploy them when the time is right.