Libertarian Utopias Require Boatloads of Computation

The first time I heard the idea that roads should be privatized, I had the reaction most people do: “How on earth would that even be possible?”

“You’d need to have cameras all over the place, and each of these companies would have to keep track of how far you drove on which road, at what time, and then add up this bunch of tiny fees, and then you’d, what? Get a monthly bill?  All of those cameras would be incredibly expensive, and you’d need to pay people to go through the film afterwards to figure out which licenses plates were which, and you’d still probably miss some. On top of all that, there’d be no way to know which roads were cheaper in advance of your trip. People won’t want to call some 800 number every time they drive, just to get the cheapest route today. So the idea of prices mitigating traffic makes no sense. It’s got to be cheaper to just have the state build roads, than to have a bunch of separate private businesses try to do all this work of tracking the movement of so many people.”

You might ask why I imagined human beings looking at film, to figure out which cars went where, instead of the obvious solution of machine learning. Why did I imagine an 800 (toll-free, for you non-americans) number instead of a mobile app?  The answer is that I first heard the idea of ‘privatized roads’ some time maybe in the mid 1990’s. This was before I was aware that the internet existed, so the thought of ‘putting your route into mapping software that includes prices as a parameter’ didn’t occur to me.  Neither did the idea that computers with cameras would be eventually be dirt cheap and ubiquitous, nor the idea that you could have computers scan through this video much more cheaply than paying people do it.  

My political beliefs were heavily framed by the current technological level of my day.  Film, cameras, and 1-800 numbers existed when I first heard about privatized roads. The internet, machine learning, and mobile apps did not. 

I didn’t bother to evaluate whether or not privatized roads were a good idea because they seemed totally infeasible back then. And they were. Now they are no longer infeasible at all. It’s easy to imagine how a privatized road system would work.  Instead of being a question of possibility, privatized roads have moved to a question of whether they are a good idea. Of course, it’s still a question few people are asking, because nobody likes libertarians, but whatever.   The difference between this system being “totally impossible” and “doable, but of questionable value” is the presence of lots of computing technology.

Every single libertarian ideal has the same basic shape: you need a bunch of computing power for it to be feasible.  Likewise, most arguments against libertarianism are arguments that the system won’t work as described – not that it would be undesirable if it did.

Tracking Reputation

For example, you might be able to summarize libertarian positions on legislation and regulation as “reputation works effectively to prevent bad actors.”  The degree to which a person believes this is probably heavily correlated with the extent of their libertarian beliefs. Look at how services like AirBNB and Lyft depend on their reputation engines. These systems look more ‘libertarian’ than the models they disrupted, because licensing and a state mandate to do business were replaced with a reputation-based model which relies on everyone carrying a mobile computer around at all times.


You might view the cab medallion system as a centralized source of truth as to who is a good cab driver.  The state does this computation, and pronounces its solution as correct. If you didn’t have computers tracking all of these things, how would you possibly build a system like lyft or AirBnB?

Arguments against Lyft, Uber, and AirBNB point to awful experiences (and there are plenty of them to go around), as arguments that these reputation systems don’t actually work.  It’s hard to imagine anyone objecting to these services if they believed the reputation systems worked perfectly. The political argument ends up being an argument about the correctness of the reputation algorithm. Arguments that these systems are undesirable usually boil down to claims that bad actors can and do have good reputations.

Tracking Negative Externalities

Another example here is preventing negative externalities.  A libertarian take on preventing pollution might go something like this: The best way to prevent pollution is to allow consumers to make the choice to buy products that don’t contribute pollution, over products that do.  At present, this is almost impossible to do, for reasons that are entirely computational.  

By enacting pollution controls via legislation, a government acts as a single source of truth on what’s safe for the environment, and what’s not. In a privatized system, you’d ostensibly have a bunch of actors computing safety levels of various chemicals, with consumers integrating this information into their purchases. It’s impossible to imagine that working a hundred years ago; even now it sounds like a stretch. Maybe in the future it will be possible, but only with a bunch more computing power, and technology to allow people to combine information from multiple sources of various levels of trustworthiness.

Again, arguments against libertarianism take the form of saying that the distributed computational system for solving the problem wouldn’t actually work.  In this case, a non-liberatarian would argue that businesses will find lots of ways to pollute that the privatized systems couldn’t track, that polluters would lie and the truth-combining software wouldn’t catch them lying, and that lots of consumers would buy cheaper goods that caused pollution, leading to a free rider problem that no reputational system would catch and punish.

So the question “is libertarianism a good idea” ends up being an argument about the computational tractability of problems in distributed computing. Is a single centralized agency with one big set of teeth more correct at assigning liability for negative externalities, than a number of distributed agencies informing lots of consumers, with the tiny teeth of individual choice-making?

Another libetarian take on pollution might say that anyone who emits pollution into the atmosphere is doing damage to my body, which is my personal property.  In order to hold them accountable, I’d need to get a bunch of people who suffered the same damage, and take the polluters to court, which brings us to…

Coordinating Responses to Bad Actors

Here’s a pattern you’ll see all over the place, once you start looking for it.  A bad actor does a small bit of harm to a large, dispersed group of people, in order to provide a lot of benefits to a much smaller group of people. The larger group being harmed either is unable to or unwilling to bear the costs of coordinating themselves to stop the smaller group doing the harm. Pollution is a great example of this. Think of the pollution from a single factory. The people who are harmed are harmed so little they might not even realize it, while the people who are benefitting benefit so much that they will gladly pay the costs to organize and defend their gains.

Coordinating activity at scale is hard, largely for computational reasons. I’d have to think about this problem. I spend most of my time thinking about my daily life, and I still feel like I don’t have enough time to take care of my immediate family in the ways I’d like. My my personal gain is almost certainly outweighed by the cost I’d personally have to expend.  The current system of solving this problem is a class action lawsuit, which is large, expensive, and usually benefits the lawyers more than anyone else.   

If we wanted a libertarian system that were capable of stopping companies from polluting, we’d need to make it extremely cheap for people to coordinate defection against bad actors. This is only feasible with a bunch of computing done around each purchase.

And, again,  the argument against libertarianism here is not that it’s desired outcome is bad, but that it won’t work, that bad actors can still get through.  In short, if you think reputation is a solvable problem, it’s much easier to see libertarianism as a good idea, with good behavior coming from desire to cultivate a positive reputation. If you think reputation is an unsolvable problem, you probably believe we must rely on the government to punish bad actors, creating an incentive to behave correctly.  

And thus, without realizing it, most people’s political beliefs are heavily influenced by their belief in the computational tractability of a game theory problem: “should I defect against, or cooperate with, this stranger?” Given that our species’ computational capacity increases every year, it seems reasonable to me to consider that political structures which were unimaginable a century ago are now eminently reasonable.

Does it sound weird to argue that libertarianism is computationally expensive? To most people, sure.  But when you remember that human beings are computers, this all fits together. Of course different forms of information technology will enable different governments. You couldn’t have any government at all without writing. Writing is the original technique humans invented for storing information outside their minds. Those fancy databases everyone’s been talking about were originally invented in sumeria, thousands of years ago. Instead of poking clay tablets with a stylus, we poke metal disks with magnets. 

Every new system of information technology enables the possibility of new forms of government which weren’t previously feasible.  Just as modern democracies require mass media such as the printing press and broadcast communications, you can’t have a functioning libertarian system without boatloads of computing power.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.