STATism: An Alternative to Democracy

I want to feel young again, so I designed a wildly unrealistic form of government that will likely never be implemented. Democracies and democratic republics (“democracies”) have been around for thousands of years. Counting the votes for each side isn’t a bad way to decide who should have power, but with the latest advances in math and machine learning, we should be able to do better. To make things confusing for libertarians, I’m going to call my system “STATism.”

I encourage you to read this as political science fiction rather than as a serious proposal.

Democracy: Advantages and Disadvantages

Ideally, any successor to democracy should have all the benefits with less of the drawbacks. It’s easy to find fault with any system, but Winston Churchill dubbed democracy the “best ever tried” for a reason.

Here are some of the advantages of democracy over a monarchy.

  • Democracy serves as a check on power. A leader has to consider the anger of an electorate before taking any potentially oppressive actions.
  • Democracy transfers information. Voters are more likely to support politicians who promise to enact policies they think they will like.
  • Democracy makes people feel empowered and important. Feelings are important, especially when they can help prevent costly revolutions.

But, of course, it’s even easier to come up with a list of problems with democracy.

  1. Money corrupts the process, distorting information. When I moved from Louisiana to Texas, I was shocked to learn that liquor can’t be sold at Wal-Mart or CVS. I’ve never met anyone who supports these policies, but I can imagine well to-do liquor store owners influencing local elections. More dramatically, our nation spends billions subsidizing unhealthy food and building tanks that will never be used.
  2. Voters are required to speak on technical questions like climate change or economics. Although citizens may be quite competent in their private life, it is unreasonable to expect them to be experts on every subject. Bryan Caplan has shown that voters systematically disagree with professional economists. Most voters are productive citizens who specialize in whatever their career is. It’s unreasonable to also expect them to specialize in wonky political topics.
  3. The “will of the people” is measured noisily. People usually don’t get to vote on individual issues. Instead, they have to pick from a limited number of policy portfolios. If you want less economic regulations but more spending on infrastructure, you cannot express your wishes in the voting booth.
  4. But it gets worse than that! Instead of noisy information about what policies the electorate supports, elections often devolve into messy signaling contests. Rather than thinking with the part of their brain that judges policies rationally and reasonably, citizens are encouraged to show off their intra-country allegiances. Most votes boil down to “I’m not like those unloyal, uncaring, selfish, misguided, godless, backwards, barbaric, pansy, prius-driving, gas-guzzling goons. I care about my country!” At best, this transfers little to no information. At worst, this lowers cultural unity in a social war of all against all.
  5. Leaders have bad incentives. In valuation terms, governments are subject to changing discount rates. A recent governor of Louisiana spent $6.9 million building a parking lot in prime downtown property. The building earned between $300,000 and $400,000 per year. Up against term limits and budget pressure, her successor leased the garage for 91 years for $2 million. This is likely a bad deal long-term, but the current governor has no reason to care. Say what you will about monarchies, but at least kings and queens (theoretically) want their children to inherit a successful country. A democratic ruler knows his reign is temporary and has an incentive to get as much from it as he can.
  6. Illinois_District_4_2004Geographic regions are arbitrary or lead to gerrymandering. In America, the party in office can redraw election districts at will. Imagine a state with three districts all evenly split between two parties. Now imagine that party A redraws the lines to that 70% of the first two districts always vote for party A and 60% of the third district will always vote for party B. Party A will never be significantly challenged in the first two districts, and it will win every state-wide vote by a margin of 2-1. Both major parties engage in this behavior, and the result is national legislators pandering to their extremists in their own elections in order to turn out a rigged vote. We are left with a set of candidates who only know how to speak to their supporters rather than reaching across party lines. Of course, one could design a democratic system which doesn’t allow for gerrymandering, but over decades of growth, even reasonable divisions can become arbitrary. Should North California really be included with South California? Or Miami with the Florida panhandle?
  7. Ken Arrow’s Impossibility Theorem states that it is impossible to develop a democratic system which satisfies certain “fairness” criteria. This is a fairly technical point and this section is already longer than intended, but if you read the fairness criteria individually you’ll say “that’s sounds important” about all but one of them. Our current system doesn’t come close to satisfying even the fairness criteria which sound important. In non-technical terms, Ralph Nader and Ross Perot can “spoil an election” by stealing votes from the party they’re more like. This gives voters incentives to vote for candidates they dislike to keep from “wasting their vote.”

All of these aren’t complaints about a particular democratic system. They’re fundamental problems with democracy.

We’re tempted to treat the latest developments as inevitable rather than fashionable. The past was weird. The distant past was very weird. Things have gotten increasingly normal, and one could continue that trend and predict things will stay normal forever.

But imagine yourself as a well-to-do European in the middle ages listening to the idea of democracy as a weird thing the Greeks and Romans used to do. You’d likely react with superior bemusement if someone said to replace the king with civilized mob rule. Democracy isn’t “the end of history.” It’s where we are at the moment, but it isn’t necessarily where we’ll be forever.

And beyond that, it has obvious downsides. I’m convinced it’s possible to do better.

But before I introduce STATism, let’s learn about a machine learning algorithm.


Suppose you had to understand a million pages of text. You could read them all, but that would take a long time. Instead, you decide to write a computer program that determines which sentences are the most important to read.

Believe it or not, this is a solved problem. There are machine learning algorithms which reduce a dataset to its most representative “exemplary” elements. One brute-force approach is to define a “distance” between any two sentences based on an automatic reading of dictionaries, thesauruses and the words used. Then try sets of K exemplars until you’ve minimized the distance between exemplars and the points they represent.

This doesn’t scale well to large data sets, but there are some clever ways to approach this problem, including one dubbed Affinity Propagation. As an added bonus, some of these algorithms will automatically select the optimal number of exemplars.

The Science article which described Affinity Propagation ended with an appendix which chose four exemplar sentences to summarize the entire paper. I had mindlessly highlighted two sentences while skimming the paper, and these both ended up being selected by the algorithm: “Affinity propagation identifies exemplars by recursively sending real-valued messages between pairs of data points” and “The number of detected exemplars (number of clusters) is influenced by the values of the input preferences, but also emerges from the message-passing procedure.”

Exemplar selection falls under the general category of “clustering algorithms.” They can also be thought of as compression algorithms, since the data is compressed to fewer important data points. This can be used for computer vision, computational biology and maybe even:

STATism: A Proposal

Suppose that every two years every citizen fills out a multiple choice questionnaire about their political beliefs. Their answers, along with demographic information such as age, income and location, are combined into a distance formula that measures how similar any two citizens are to each other. Then, exemplar citizens are selected to fill the legislature for the next two years, in the same way that juries are asked to decide trials. Point by point, this system is far superior to traditional democracy.

  1. Money doesn’t corrupt, since no one knows who to bribe. People don’t even “run” for office. They just happen to mathematically represent their peers well at that period of time.
  2. Exemplars don’t come up for re-election, so they don’t need to spend time fundraising, trading favors or appeasing voters. Hopefully they’ll take their job seriously and respect technical questions.
  3. The will of the people is measured precisely on individual questions rather than through messy policy portfolios.
  4. Votes don’t devolve into signaling contests since all questions are questions of policy. You promote your policy by discussing your policy and changing how you vote.
  5. Exemplars owe no one anything for their power, so they have less incentive to be corrupt than traditional politicians.
  6. There are no geographic regions to gerrymander.
  7. The Ken Arrow critiques are irrelevant.

At this point, some readers not fully into the spirit of this silly paper will start questioning the details.

  1. Where will the questions come from? I don’t know, maybe people could submit them OKCupid style. The algorithm handles blank data, so fully completing the survey isn’t necessarily necessary. This means the surveys can be arbitrarily long.
  2. How will exemplars pay the bills? I don’t know. Pay them the greater of their projected income plus 20% or the country’s median wage. Make it worth it to be an exemplar and celebrate if an expensive person takes up the mantle. Yeah, that sounds expensive, but it’s probably cheaper than the U.S. status quo where the most junior representatives make $174,000 annually, not counting benefits. It’s probably far, far cheaper than the U.S. status quo where well-connected factories get to produce tanks that are never used.
  3. Are we sure two years is the optimal term length? Of course not. Are you sure that the status quo Senate should have 100 senators? Of course not. These are hard parameters to tune, and I don’t claim to have perfected the approach.

Yes, there are details which need fine tuning, but that the challenge of democracy is fundamentally a data compression problem, and we could use the latest machine learning techniques to solve it optimally. Having people choose between two arbitrary parties seems like a bad solution.

STATism: Potential Failures

I’m somewhat attached to this idea since it is my own, but I’m not convinced it would be a positive change.

It’s possible that the failures of democracy are blessings in disguise. Status quo representatives are naturally from the upper crust, and maybe it wouldn’t be sustainable to optimally have representation from all classes. Maybe democracy doesn’t result in people getting what they want, and that’s a good thing.

I don’t have an answer to this except to acknowledge that it’s a possibility.

Another possibility is that this will make the “election” process boring and no one will want to participate. Good, I say. The election process should be boring. It’s a serious matter, and you shouldn’t get to co-opt it for your silly signaling games.

Some may wonder how we would elect the executive branch since I’ve only specified how to elect the legislative branch. I don’t want to answer this question. Maybe the exemplars should have a vote amongst themselves. Maybe there should be a convoluted “electoral college” system. Maybe there shouldn’t be an elected executive branch at all. Maybe the head of operations should be hired by the exemplars like any other CEO. I have no clue. I’m just some guy.

All things considered, this system seems to accomplish the stated goals of democracy more effectively than traditional democracy.

How to Make STATism a Reality

You can’t.


No buts. You should probably focus on your career instead.

I suppose you could take STATism as a criticism of modern politics and the democratic process. To me, that means not taking political parties too seriously, looking for common ground and listening to your political opponents with patience, charity and curiosity. You could focus on your local environment and foster respectful and fair conversations.

Worst case scenario, if we ever need to start over again, keep this proposal in mind.