What is Power?

And can we reason about it precisely?

When I was younger, a friend and I were driving out of New York City and headed upstate at night. It was a 3 or 4 hour drive. Unsurprisingly, a police car pulled us over though we weren’t speeding.

The officers separated us. One of us had to stay in the car while the other was taken out and questioned: did we have anything on us, was there anything in the car, did the other one have anything. Then they switched us and asked the same questions again. They were trying to get one of us to give up the other.

Then they searched the car; even though we hadn’t consented. They didn’t find anything, they didn’t give us a ticket. They just told us to get out of there.

This wasn’t my first or my last encounter with the police but it was the only one on a road late at night with no witnesses. So while they were searching the car, I remember being worried that they would plant something. That’s not an abstract fear; police plant evidence often enough that worrying about it was rational, not paranoid. But what was worrying me wasn’t just that “something” could appear in the car, it was that if something did appear, we’d have no way to contest it and that our lives could be basically over in that moment. It was really unsettling. The officers write the reports and their account is the account the criminal justice system uses. It becomes, for practical purposes, the truth.

In this situation, I had agency. I could answer or refuse to answer or assert my rights. I could stay calm or argue. I could say that I didn’t consent to the search. But the parts of the encounter that really mattered weren’t mine to set. The officers decided whether we were stopped, whether we were separated, what counts as suspicious, whether the car was searched, when we could leave and what the official record is.

There’s a difference between having choices in a given situation and controlling the situation. This distinction is important.

Crypto-for-Something

I’m a cryptographer by training, but for a long time I couldn’t have told you exactly why. I write cryptography papers, I review for and chair cryptography conferences, but I don’t feel about cryptography the way some of my colleagues do. I like it, but I’ve met people who love it, who think about nothing else and want to think about nothing else. I’m not one of them. I view cryptography as a tool and not as an end in of itself, so I clearly don’t do crypto-for-crypto. But I do crypto-for-something, and for a long time I couldn’t say what that something was.

The same is true of the rest of my work. At different times I’ve worked on security, privacy and anti-surveillance. Surveillance and privacy both matter to me, but in a different way than they matter to many of my colleagues. And it’s not just security, cryptography and privacy. I circle around subjects with no obvious connection, like the law, political science, game theory and algorithmic fairness.

Over time, I realized that what interests me is something underneath all of these domains. What really interests me is power, but I have never had a precise way to talk about it. So I’ve been thinking about power a lot and trying to articulate the intuitions I have more formally. This isn’t an open problem asked by anyone but it is something I need to do in order to clarify my own thinking.

Phil Rogaway, in his essay The Moral Character of Cryptographic Work, argued that cryptography rearranges power and that cryptographers are responsible for how it does so. I agree, and I’d put it more strongly. To me, cryptography is interesting mostly to the extent that it rearranges power. Or, put another way, what I care about is not crypto-for-crypto or crypto-for-privacy or crypto-for-security but, more fundamentally, crypto-as-power and I would like a way to articulate that precisely. The same goes for my other domains of interest. The law, game theory and fairness hold my attention because at some level they’re about who has power and who doesn’t.

What I mean by power. When I say power I don’t mean it in the loose sense of someone being important or influential. I mean something more specific: control over the systems in which other people act. Consider a court trial. You can hire the best lawyer, mount the strongest defense and conduct yourself perfectly in court, and still your fate isn’t settled by your actions alone. Someone wrote the law you’re tried under. Someone decided what counts as a crime, set the mandatory minimum and fixed what can and can’t be admitted as evidence. Again, there’s a difference between acting within a system and controlling the system itself, and it’s the latter that I mean by power.

Theories of Power

Power is difficult to pin down. It’s been studied for as long as there’s been political thought by scholars in different fields.

Max Weber was a German sociologist who is credited with co-inventing sociology around 1900. In his work Economy and Society, he first distinguished between raw power and authority. In his framework, raw power is the ability to make you do something, by force if needed. Authority, on the other hand, is the ability to make people obey because they accept that you have the right to command them. For example, a manager, a judge and the state all have authority. Weber wanted to understand why that acceptance exists, whether people obey out of tradition, or out of loyalty to a charismatic leader, or out of respect for rules and offices. Note that the people in charge are visibly in charge, so this is a form of power you can see.

Robert Dahl was an American political scientist who, in the 1950s, tried to turn power into something you could measure. His definition is the one many people come up with on their own: A has power over B to the extent that A can get B to do something B wouldn’t otherwise do. A feature of this definition is that you can observe it, e.g., in politics or in a courtroom, by studying which party wanted which outcome and seeing who prevailed. Dahl did exactly this by studying a series of actual decisions in the city of New Haven and found that power was more spread out than people assumed, with different groups prevailing on different issues. This kind of power is still visible because there is an open contest with a visible winner.

Steven Lukes is a British sociologist who, in 1974, wrote a book called Power: A Radical View. His argument was that the open contests that Dahl wanted to measure only reflect the first of three faces of power. The second face is quieter and is about setting the agenda. In other words, power isn’t only about winning a contest, it can be about making sure the contest never even comes up. In Lukes’ framework, the third face of power is the ability to shape what people want in the first place so a conflict is never even created. If you can get someone to see their situation as normal, or natural, or just the way things are then you don’t have to win anything because there is nothing to fight about. That’s the deepest kind of power and the hardest to see because from the outside it looks like consent.

Michel Foucault was a French philosopher who proposed a different conceptualization of power. For Foucault, power isn’t something a person or a group has. It is spread through everything, schools, clinics, prisons, exams and it works less by stopping you than by quietly shaping what you do. His best-known concept is the panopticon, an eighteenth-century prison design where one guard in a central tower can see into every cell but no prisoner can tell whether they’re being watched. The consequence is that they behave as if they are always watched and they start policing themselves. Foucault thought modern life worked this way and that power is most effective when it’s everywhere and nowhere, when nobody has to be watching because you’ve already learned to watch yourself.

Notice a few things. These accounts are general in the sense that they are not restricted to a particular domain like economics, sociology or voting. They are informal and, more interestingly, they all define power differently.

Power has also been studied in game theory and economics. But these accounts go the other way: they are formal but focus on a kind of power that is specific to a domain and problem setting.

I’ll come back to all these ideas in later posts because each of them is interesting in its own right.

Power Systems

Social and socio-technical processes like traffic stops, elections and hiring processes are complex systems whose power dynamics I would like to understand better. So the starting point is to model them with what I call a power system, which includes a process together with a power structure. The process is how the system runs and the power structure determines who is allowed to change which part of the process. Note that “change” here is concrete. If an agent holds a power, then it has the authority to replace one component of the process with a different component chosen from some set of alternatives. You don’t hold power in the abstract, you hold the ability to swap out one part of the system for another.

What are these components? They include the components that determine who can participate in the system, what actions people are allowed to take, what the consequences of those actions are and what people can see. They also include the component that determines who can control what.

In the encounter I described above, the officers, my friend and I were participating in a process. But it was the officers that decided our admission into this process, not us, based on their suspicion. The officers controlled what counted as suspicious and the consequences that would result from the narrow set of actions at our disposal. They also controlled what we could see and know. They split us up so neither of us heard what the other said, and they were the ones writing the report.

The reason to separate the process from the power structure is that one doesn’t determine the other. The same process can run under completely different arrangements of who controls what, so two systems can run identically, reach the same outcome and still differ entirely in terms of the power structure. Power is also a matter of what could be changed and not only of what actually happened. To understand what power someone has, you can’t just think about what they do inside a fixed system. You have to reason about the range of alternative systems they can produce by changing the parts of the system they control.

The officers who stopped us could have let us go, searched the car, charged us with something or written almost anything into the report. But those are actions, things they could do during the encounter. The power they held was exerted even before the encounter started and included their ability to stop us in the first place, their ability to carry weapons and the legitimacy of their word over ours. And if they had planted something, the stop would have ended and we would have entered a new system, the criminal justice system. In that system, we would have been defendants and the officers witnesses.

Moving Upstream

Systems feed into other systems and the ability to control one can shape what happens in a downstream system. So it’s worth asking what sits furthest upstream. The criminal justice system we could have entered doesn’t run itself. Its laws and the people who enforce them trace back to elected officials and those trace back to who gets to vote because in a constitutional democracy, the vote sits upstream of many socially important systems. So whoever controls the vote controls, eventually, what can be done to people during a stop. The civil rights movement understood this well. Black people in America had the right to vote since 1870, when the Fifteenth Amendment was ratified. But the Southern states spent most of the next 100 years making sure they couldn’t use it. The states controlled who could register and used various mechanisms like literacy tests, poll taxes and registrars free to fail anyone they chose to keep Black voters off the rolls.

The goal of the movement wasn’t only about changing individual people’s feelings towards Black people; it was about acquiring a commensurate amount of power and, in a democracy, that goes through the ability to vote. Because of this, the Voting Rights Act of 1965 was a landmark achievement and a rare check on the Southern states’ power. Its strongest provision, preclearance, forced the states with the worst records to get federal approval before changing any election rule, splitting a power they had always held alone.

Unfortunately, the Supreme Court has spent the last decade trying to dismantle the Voting Rights Act. It struck down the coverage formula in Shelby County v. Holder in 2013 and recently---while I was thinking about all this---decided to severely limit when race can be used to draw districts, in Louisiana v. Callais. Neither ruling is about counting votes. They are about setting the rules and both hand control of those rules back to the states, with predictable consequences: several have already used it to pass new voter ID laws, close polling places and purge voter rolls.

What is the Point?

A criticism this kind of work often receives---and that I have already received---is that the payoff is not clear. That once you get past all the definitional machinery, it feels like a vocabulary for things people already understood so what was really accomplished?

I think this misses the point and it misses it in a way worth being explicit about. It measures the work (usually) by how hard the math is, but hard math is not the goal. The payoff of a framework is precision and clarity. It is in how it changes the way you think and see things, not in how clever the math is. What you could only hand-wave before becomes something you can state, compare and argue about. It is in how it allows you to see that the same few ideas can reach across completely disparate fields that don’t usually talk to each other.

This post ran a traffic stop, a court trial and voting through one vocabulary, one set of concepts and traced power upstream and down between them. That’s the payoff and, in my experience, this is how useful ideas start. So the right question isn’t whether the math is interesting, but whether the framework allows us to say things and ask questions we couldn’t say or ask before.

Future Posts

This is the first in a series of posts I will make on power. As I mentioned, I’ve been thinking about these ideas for a while and they have helped me clarify my own thinking about cryptography, surveillance, AI, voting rights, democracy and algorithmic fairness. Hopefully they can help clarify other people’s thinking as well.