Some technical arguments against tyranny
Distributed systems and economics are instructive, if not definitive
Historically, the most common form of human governance has been some form of monarchy, often shading into tyranny. With that perspective, it's perhaps not all that surprising if we currently are finding it difficult to avoid a transition into tyranny.
The founders of the United States were very concerned about avoiding tyranny, in ways that were sensible for their time. Indeed, much of the Declaration of Independence consists of reciting the unacceptably tyrannical behavior of George III. Some of those complaints may have an oddly familiar quality for readers who are familiar with current events:
“He has endeavoured to prevent the population of these States; for that purpose obstructing the Laws for Naturalization of Foreigners”
“He has erected a multitude of New Offices, and sent hither swarms of Officers to harrass our people, and eat out their substance”
“He has kept among us, in times of peace, Standing Armies without the Consent of our legislatures.”
“He has affected to render the Military independent of and superior to the Civil power.”
“For Quartering large bodies of armed troops among us”
“For cutting off our Trade with all parts of the world”
“For depriving us in many cases, of the benefits of Trial by Jury”
“For transporting us beyond Seas to be tried for pretended offences”
Further insight came with the experience of the 20th century. One important source is Hannah Arendt’s The Origins of Totalitarianism. For anyone who is interested in avoiding tyranny and/or totalitarianism, a number of valuable resources are available; a failure to engage with these issues must be understood as a choice, a failure of will, rather than as a lack of available knowledge.
But it may be that it's hard to get the bulk of people to understand the significance of avoiding tyranny, if it hasn't been a real threat in recent memory. And for those who are sympathetic to a would-be tyrant’s goals, it may well seem as though installing one as tyrant would be a net positive. Perhaps they see democracy as mere pious platitudes, repeated by many but believed by none.
A cynical view might be that democracy is just a story told by elites to distract from their misdeeds. If someone were indeed in the grip of such a full-on disbelief in the merits of democracy and the rule of law, what evidence beyond “tradition” or “my personal preference” could we muster?
A technical approach
We now have decades of experience with building large distributed systems, which provide the underlying implementation for pretty much any modern national or global business. Those systems involve forms of cooperation among autonomous entities, but they are designed and controlled in ways that human societies aren’t. What you can and can’t accomplish in a distributed system doesn’t determine what you can and can’t accomplish in a society, but it’s certainly a set of constraints that is useful to understand.
Although I don’t expect to convince anyone who favors our current slide toward tyranny, I think it’s worth sketching some technical arguments for the merits of democratic societies and the rule of law, based on distributed systems and economics. (In some ways this piece is a continuation of what I have previously written about the rule of law and reliability, about the nature of rule of law, and understanding rule of law as a mediating format).
I think there is evidence for the following propositions:
Elections are important
Meaningful democratic influence is important
Rule of law is important
It’s worth noting that any one of these, considered separately, can be tackled in a somewhat cynical and low-impact way. However, putting all three together is a combination that aligns well with what constituted conventional political assumptions in the US before 2025.
1. Elections are important
A leader is better than direct democracy
An important, if awkward, observation is that systems arguments can be understood as supporting monarchy over democracy, in some limited ways. Intellectual honesty requires underscoring that, rather than just claiming boldly that “experience with computer systems supports democracy,” or similar. If there is a set of otherwise identical participants, it is generally more efficient to have one of the participants designated as leader. The mechanism for leader selection need not have any relationship to democracy or to fault tolerance. It’s simply that it’s generally simpler, more efficient, and more robust to have a single participant make certain choices and impose them on the collection, rather than having everything determined by the group collectively.
We can see a rough counterpart in civic governance in New England, where only the very smallest towns even attempt to implement anything resembling direct democracy. As a town grows in population, it invariably moves away from direct democracy toward some combination of representative democracy and professional management. Typical entities that materialize along the way are: a representative town meeting to replace a direct town meeting; a town manager to be a single unelected executive; a select board to be a small-group elected executive; or a mayor to be a single elected executive. In all these structures, the electorate reduces its burden of votes and actions, delegating some or all of that work to elected or unelected specialists.
Leader replacement is required for robustness
A single leader is efficient, but potentially constitutes a single point of failure. Accordingly, systems with leaders must have some kind of leader selection mechanism if the system is to be robust. Building computer systems offers the insight that this selection mechanism needs to be fairly robust in itself, or else it can become the point at which the system fails. (As people attack multiple aspects of our current political election system, often under the Orwellian terminology of “securing elections,” this insight is important to keep in mind.)
The pragmatics of building a system don’t require the selection process to be fair or democratic or representative. It’s fine to always pick the same element as leader, as long as it can do the job. (Of course, depending on the technologies involved, it may be wise to share the leader role roughly evenly, to avoid any possible concerns about “hot spots” or uneven wear).
The one and only point of a leader selection process, technically, is to have a single unambiguous locus for control and decisions. This turns out to be surprisingly hard in practice. Although there are mechanisms that work correctly almost all the time, people who run large distributed systems still encounter rare “split brain” situations, in which two or more components have simultaneously declared themselves to be in charge.
2. Meaningful democratic influence is important
Distributed is harder, but usually better, than centralized
Friedrich Hayek made an argument in The Road to Serfdom that central planning couldn't work in situations where free markets could. The main idea is that the multiple independent players and multiple independent negotiations happening in a free market are inherently more flexible and better informed than a central planner attempting to aggregate the same kind of information and process it. A consequence of the weakness of central planning is that all efforts at planning devolve inevitably to some form of dictatorship, which was Hayek’s key warning: essentially, don’t imagine that you can have a planned economy coexisting with democracy, because effective planning requires central control.
For our hypothetical “OK with tyranny” audience, the incompatibility with democracy wouldn’t necessarily be a problem; but we can invert Hayek’s analysis to say that dictatorship wouldn’t be as effective at making necessary economic tradeoffs. Even people who think they want and would benefit from a tyranny would find that the centrally planned economy would usually not match their preferences. The tyrant’s central control would dominate all the choices that would ordinarily be individual choices, even for those who think of themselves as supporters of the tyrant.
Do computers overrule Hayek?
The temptation here is to argue for “a really big computer” or “lots of data” as the magic solution: if we just make the central planning fast enough, smart enough, and with enough accurate inputs, then the tyrant can get central planning to work. But the reality is that distributed systems can do things that centralized systems can’t, partly based on physical laws of the universe like the speed of light.
The overall argument for the superiority of distributed sensors and distributed decision-making is (of course) familiar to those of us who have worked on distributed systems for decades. The ubiquitous example of a distributed system is the internet, which is simultaneously “always broken somewhere” but also “never broken everywhere at the same time.” (Such “partial failure” is characteristic of distributed systems, and helps to distinguish them from centralized systems.) Installing a central global controller for the internet is something that a tyrant might attempt, but that would be unlikely to succeed; the nature of the global internet is that it routes around damage and restrictions. That said, multiple tyrants around the world have reinforced their tyranny with their own local national restrictions on the internet.
In general, centralized systems are easier to build and easier to keep running, but harder to scale up, harder to make responsive, harder to make fault-tolerant. The merits of distributed systems become apparent when we consider the idea of organizing a society, particularly when we consider organizing a “high end” wealthy society: one with substantial innovation in an unpredictable environment, with high complexity of societal organization in terms of both the structures of companies and the structures of industries.
Distributed systems are like democracies
We can draw a rough parallel: monarchy vs. democracy is roughly comparable to centralized vs. distributed systems. Distributed systems involve smaller corrections, better access to local information, and less dependency on correct choices by some central authority.
In theory, going back to Plato, the ideal ruler is a benevolent dictator, a philosopher king. The problem is that it's not clear such a creature exists; or, at least, there doesn't seem to be a very high frequency of them, and there doesn't seem to be any reliable way of making sure that they are in charge. For example, we could consider Marcus Aurelius to be an example of a relatively well-behaved Roman emperor, who was in his own way a philosopher. Certainly, he was more thoughtful about his life than seems to have been true of Roman emperors in general. But then we must step back and consider that his reign was only a small number of years in the entire history of the Roman Empire. Meanwhile, there were dozens of other rulers who were collectively horrible. In addition to the low quality of the rulers, the transition between rulers was frequently chaotic and bloody. Even if we can’t necessarily promise better quality rulers via democracy, we could hope for cleaner and more peaceful transfers of power.
What we can really expect from a monarchy is not superior order compared to a democracy, or superior lives compared to a democracy. In particular, the messiness and infighting of democratic processes do not go away in a monarchy – they just move to the royal court and become contests among courtiers.
3. Rule of law is important
Relationship to rule of law
Let’s consider a tyranny’s decision-making processes. Instead of a relatively predictable rule of law, the decision process comes down to individual personal choice by the tyrant or by the tyrant’s ministers. That unpredictability makes it harder to make substantial investments, and so everything necessarily becomes smaller scale and more cautious.
It’s worth noting that the tyrant can’t bully their way out of this phenomenon. Even if they want people to invest more and take more risks, their ability to affect these decisions is limited by (a) the number of people they can personally bully and (b) the ready availability of many different reasonable-seeming explanations for reduced investment, other than the tyrant’s behavior or the absence of rule of law. If a person wants to avoid investment at the level the tyrant wants, they don’t have to admit that the tyrant is the problem.
There is an illusory version of “taking charge” of all investment decisions, perhaps by centralizing all capital under the control of the tyrant, but that just returns us to the centralized vs. distributed economic issues discussed in the previous section.
The decline of rule of law leads either to reduced investment (if distributed decisions are still allowed) or to poor quality investment (if decisions are centralized). Those who want a tyranny and imagine that they will get a thriving economy are almost certainly incorrect, in general.
(Of course, it’s possible to construct exceptional cases. For example, perhaps the economy is hobbled by a very specific self-inflicted problem that the tyrant removes. That’s a case where the advent of tyranny might improve the economic situation. But in general, rule of law is not only more pleasant but more valuable economically.)
Indeed, as I’ve written previously (and following Hayek), predictability is fundamental to the rule of law. Even with this predictability, one can’t necessarily predict the outcome of a particular dispute or tradeoff, but one can predict the process for making the decision. If decision processes are arbitrary, that is not rule of law.
This means that rule of law is comfortable territory for computer scientists, distinguishing systematic processes from chaotic or arbitrary processes. The rule of law is thus a cousin to algorithms.
(We should be careful not to overstate the similarity: there is a long history of dubious efforts to render legal judgments via computer programs. We are likely heading into another phase of such misplaced enthusiasm, fueled this time by large language models (LLMs).)
Nevertheless, we can say that there is something important in rule of law that is like programming: a rough consensus among participants that as the process reaches its decisions, the outcome is unrelated to the specific persons involved.
Legal, but not rule of law
Some recent activity on tariffs is an example of something that may well be “legal” but is nevertheless “contrary to the rule of law.” The details are somewhat complex and currently in litigation (related to emergency powers and whether any genuine emergency exists), but the essential point is apparent with a simplified version.
If Congress passes a law that delegates tariff rates to the president, then the president is clearly acting legally each time they set a new tariff rate. However, if the president’s decisions follow no apparent process, or if the decisions are obviously tied to specific persons, then those changes are not in line with the rule of law. A process that is “predictably” tied to the arbitrary personal choices of an individual is not really a predictable process: it is an example of “meeting the letter, but violating the spirit” of the rule of law.
Personal power vs. rule of law
In a system without rule of law, the economic harm (reduced investment) is exaggerated if the monarch is notorious for being chaotic or unpredictable. Here it’s important to observe that the incentives for the monarch are directly opposed to the incentives for the economy. The unfortunate paradox is that a monarch maximizes their power as an individual exactly by being unpredictable.
If one is predictable, decisions don't need to come to you. In contrast, if no one knows what you will say about anything, everything must come to you – particularly if the consequence of guessing wrong is dire.
From a systems design perspective, an arbitrary and unpredictable monarch is catastrophic: both the capacity and the resilience of the system are minimal, because everything must go to the monarch. However, from the narrow perspective of the monarch and their power, it's great: they can decide everything, and no-one else feels that it’s safe to make any decisions.
Survivability and rule of law
Systems with failures are more robust if elements can recover after failures and be reintegrated into the system. Systems are also more robust if obsolete or unneeded elements can be gracefully removed or replaced in way that is designed and intentional, rather than being ad hoc and improvised.
It would be unusual, and arguably counterproductive, to build systems in which a failed or obsolete element is simply destroyed. Correspondingly, we can observe that even would-be tyrants do better with rule by law. Peaceful transfers of power allow would-be leaders to try again. In contrast, transfers of power in authoritarian or totalitarian systems typically involve the forced removal of a predecessor and their execution or imprisonment. Rational would-be rulers should prefer the rule of law, since it means that they have a better chance of surviving into old age.
Unfortunately, would-be rulers seem to frequently share a personality defect in which they both overestimate their abilities and underestimate their risks. Nevertheless, a casual examination of the history of empires (and the Roman Empire in particular) establishes that being the emperor was not a situation that was conducive to a long and healthy life.