Prerequisites: Nothing strictly necessary, but I recommend knowing about Basic Income, the World Government, and Licensure to be better able to contextualize the sections on Utopia.

Kidneys are, by far, the most in-demand organ. In the USA alone, 90k people are waiting for someone to donate a kidney, nearly ten times higher than livers, the runner-up. Without a functioning kidney, people must undergo dialysis multiple days each week, hours at a time — a grueling ordeal with an average cost (per patient!) of over $100k/year (though most of those costs are paid by the public via Medicare, etc). And even with dialysis, thousands of people die each year from the kidney shortage. Receiving a kidney reduces the death rate by half, compared to others on the waitlist. A rough estimate for the total value of a kidney is a million dollars.

And you almost certainly have two sitting inside you, right there! You can definitely survive with just one! After recovering from the surgery, which is laparoscopic and not particularly dangerous,1 healthy donors experience almost no symptoms on a day-to-day basis. You might have to take it easy on certain painkillers, or other drugs. You might not do as well when severely dehydrated. Mostly it increases risk a little bit, such as risk of getting diabetes during pregnancy, having very slightly elevated rates of hypertension and kidney failure (ESKD),2 and, obviously, if you happen to get in an accident that destroys your one remaining kidney, that can be really bad if you don’t have a backup.

Still, even with the costs and risks, and only getting half of the gains from trade, I think a great many people would gladly sell one of their kidneys for $500k! That’s enough to flat-out buy a family-sized house, in many places (in the USA!), and still have money left over for a car.

Too bad it’s illegal.

It’s not just illegal in the USA, selling organs is illegal everywhere on Earth… except, randomly, Iran. And predictably, Iran is also the only place in the world where there isn’t a shortage of kidneys.

Meanwhile…

Here are some of the most dangerous jobs:

• Fishing, hunting, and logging — around 100 fatal accidents per 100k full-time-equivalent worker-years. (This is the most dangerous occupational category tracked by the US Bureau of Labor Statistics.)

• Being deployed to Iraq as a soldier — around 50 fatalities per 100k deployment-years for all roles combined, but climbing to ~200 for combat roles specifically.

• Selling sex on the street — around 200 per 100k women,3 just from murders!

That statistic comes from investigating American street prostitutes in the late 20th century, and thus might seem biased towards the high end,4 but more recent studies in Canada and the UK have found comparable levels of violence. Sex workers who make arrangements with clients online are significantly safer, but still regularly subject to violence. This 2014 meta-analysis found that close to half of sex workers across a variety of contexts have experienced physical violence as part of their job in the last twelve months.

Cracking down on prostitution increases the total harm to prostitutes, because even though harsher laws can reduce demand, they can’t come anywhere close to eliminating it. As the black market is pushed deeper into the shadows, abuses magnify, both from clients and from the pimps who are needed to enforce payment, when police aren’t an option.

Coincidentally, one of the countries with the harshest anti-prostitution laws is Iran. In addition to adultery, pornography, sodomy, and a host of other subjects, Iran treats prostitution (and pimping) as crimes punishable by death. The result? It’s hard to say with any real confidence, but in a 2022 online survey of 3,407 (urban, male) Iranians, 43 percent of respondents admitted to having paid for sex — over twice a 2008 estimate for the US rate. Wikipedia speculates that 0.28% of Iranians are prostitutes, possibly higher5 than the US.

Part of why these numbers manage to be so high, despite the harsh legal/religious/social environment, is sigheh — a temporary marriage with a fixed duration that can be on the scale of hours. Sigheh don’t require a cleric, a witness, or any documentation. A couple can technically contract one between themselves with a simple verbal exchange. Once taking the “dowry” and “matchmaker” into account, the practice serves as a thinly veiled loophole that de-facto legalizes the world’s oldest occupation.6

Iran is hardly the only country with loopholes, of course. Outside of very specific contexts in Nevada, prostitution is illegal everywhere in the USA. But being an “escort” isn’t. Pay a nice girl to show you around the motel city, and if she also happens to want to have sex… well, that’s a choice that any competent adult can freely make! Dancers, hostesses, masseuses, and other similar arrangements exist across the world, with no society in human history ever having really eliminated the practice.7

On the flip side, there is New Zealand, which fully decriminalized sex work in 2003. The government’s own five-year review concluded that the law had improved health and safety, made it easier for workers to refuse clients and to report violence to police, and — contrary to opponents’ loud predictions — had not increased the number of people working in the industry. There continue to be critics, but no rigorous study has found that decriminalization made things worse on any major metric, and the country’s sex worker collective and most independent academic reviewers consider the law a clear success. Belgium followed suit, fully decriminalizing the sex trade in 2022, and several Australian states have also moved in the same direction.

Exploitation

Selling sex and selling organs have many similarities. Part of the near-global ban on both, for instance, comes from repugnance and disgust, likely because of evolved intuitions about disease. This kind of objection is pretty foolish. We have a duty to overcome our evolved reflexes when they stop being a trustworthy guide. But there are other connections worth taking more seriously, such as concerns about cultural/social externalities around how we relate to bodies, and — perhaps most concerning — both naturally lend themselves to patterns of exploitation.

See, here's the thing: if selling kidneys was legal, the market price wouldn't be anywhere close to $500k, regardless of what the value of a kidney is to the buyer. Because each person naturally has two kidneys, supply naturally outstrips demand, leading to an enduring surplus, and competition between sellers will mean that buyers will get nearly all the gains from trade.

We can see this in the history of every place a kidney market has actually existed. In the 1990s, India was the global hub for those looking to buy kidneys. The total cost, in today's dollars, was often less than $20k, with most of that getting absorbed by the medical staff and other middle-men. Those selling their kidneys often only saw about $2500. To a poor, young Indian, that was still a lot of money — more than the wages from seven full years of labor.8 But still, the arrangement deeply favored those from wealthy countries who had enough money to be medical tourists.

Critics of the organ trade also point to bad outcomes. This 2002 study followed up with 305 Indian sellers and found that 86% reported their health had gotten worse since the operation, and 79% said they would not recommend that others sell. Average family income dropped by about a third, and three-quarters were (still) in debt. This 2001 Iranian study shows similar levels of dissatisfaction. My read of these self-reports isn’t that the loss of the kidney ruined these people’s lives per se, but rather that both societies heavily stigmatize those who sell their body, and a combination of withdrawal of social support and the continued existence of hardships9 that pushed them down that road in the first place meant that their lives continued to fall apart, despite the short-term windfall.10

This dynamic, where seller-side competition pushes the gains from trade lopsidedly toward buyers, shows up across most of the economy, from socks to semiconductors. In competitive markets, sellers retain only a thin slice of the value they help create, with net profit margins of usually less than 5%.11 Nobel Prize winner William Nordhaus famously estimated that innovators capture just 2% of the social value of their innovations, with the other 98% spilling to consumers.

In most cases, this move away from equally dividing gains from trade is broadly good. Most consumers are poorer than the owners of the corporations they buy from. Thus, when the competitive market dynamics suppress prices, this usually means there will be less money in the hands of the rich and more in those of the poor. It’s when this dynamic is flipped, such as poor workers competing to sell their labor to wealthy business owners, or farmers competing to sell their produce in stores like Walmart, that we most commonly12 see the market dynamics as bad.

Unionization can help sex workers, and better information systems, such as testimonial databases from those who previously sold and warnings about risks and hidden costs, can help reduce the chances that someone will make a deal that they’ll later regret. But ultimately, the dynamic where the poor get the short end of the stick still remains.

Federal Prohibition

One way to protect against exploitation is to make trade illegal, and try to help the poor and the sick in more wholesome ways. By banning body markets, society can also protect the sanctity of the body, reduce various forms of temptation, and generally avoid noxious externalities. I personally come down in favor of legal kidney markets, but am more open to the case for banning sex work, and recognize that not everyone will agree about how to weigh the various factors. If you are unpersuaded that there are real costs, and find yourself on the extreme libertarian end, consider whether you would endorse a market where people could sell all their bodily organs at once — effectively providing for their families by committing suicide. What might the very existence of this sort of market do to family dynamics for the ultra-poor?

But of course, taking these kinds of win-win trades off the table is easier said than done. Unless one plans to fully outlaw sex or organ donation, there will always be the loophole of “escorts” and hospital “patronage.” And with black markets come organized crime syndicates that eat away at the moral fabric of society in a thousand other ways. Fighting against what people naturally want can push societies into surveillance, corruption, and tyranny.

What are we to do, when values are in conflict across society? One option is the utilitarian logic of democracy — let the people vote, and serve the majority. If our world is any evidence, this results in compromises. Prostitution is effectively legal in most democracies, but it is heavily stigmatized and the police are allowed to harass sex workers who advertise their services too blatantly, or in the wrong places. Selling blood is allowed, but selling your eye is not.

But there is another option: federalism. Allow people to form local governments and communities that can impose whatever taboos and restrictions they decide are valid, but prohibit people from one culture from dictating what is permitted in another, even if the would-be dictators are the majority.

By encouraging a diverse palette of societies, and facilitating migration such that people can self-sort according to their priorities, it becomes possible to have a world that’s big enough to satisfy nearly everyone. Those who seek freedom can live in places with fewer laws, while those who value purity can live in communities that protect their members from temptation and exploitation.

Utopian Sex Work

My guess is that Utopia leans in the direction of having a weak world government and strong local governments. Out on the open ocean or in other remote parts of the world, there are few laws preventing consenting adults from doing whatever they want to themselves and each other. As long as prostitution, gambling, and drugs stay out of their communities, the everyday citizens of Utopia tolerate the existence of the wild outskirts as part of having a diverse and peaceful world.

In civilized areas, local governments are the norm, with most states opting for just a light touch of paternalism by requiring people to have licenses in order to engage in things that are considered risky, such as being a sex worker (or simply having sex). More intense communities, such as religious cloisters and/or those centered around raising kids, enforce strong bans against behaviors they see as taboo or sinful.13

Governments that legalize sex work almost always restrict how it can be advertised and marketed, lightly taxing it as a way of nudging people to make more wholesome choices. This tax revenue is sometimes spent on prizes for public goods, such as websites or other platforms to help sex workers share information about clients and share tips for staying safe. Other communities simply use it to encourage people to start traditional families and have children.

As a result of the variety of options, people who want to become sex workers, whether because of desperation or because they like it, usually have the resources to pick arrangements that work for them. Regardless of what community they choose, all are protected by universal laws against violence, and can always rely on police protection. And while there are some in Utopia who are desperate for cash,14 the presence of generous universal basic income (and guardians for those who need guidance) means that virtually no one is actually forced to become a prostitute in order to survive.

Utopian Body Markets

Utopians, thanks to culture and training, are less mired in evolved intuitions about revulsion and disgust. They broadly recognize that technology and medicine are responsible for improving and extending lives, and invest heavily in the future. There are large, established prizes for companies that develop ways to make new organs, such as those using 3D bioprinting with cells cultured from the host, or simply the development of implantable, artificial organs.

But these technologies are less in demand than in our world, thanks to large, well-regulated organ markets in various countries. These organ markets are somewhat controversial, due to concerns about exploitation, and are not available everywhere.15 But like with sex work, the people of Utopia don’t feel that they have authority to dictate what people in other places and cultures can and cannot do. And also similar to sex work, basic income means that people almost always have other options. Because buyers and sellers care about trust and safety, the biggest Utopian organ markets are well regulated, with clearly posted warnings about risks and health consequences, standard language for contracts, and suggested prices.

In addition to kidneys, people in Utopia can make money selling their blood (not just blood plasma16), hair, skin, bone marrow, sperm, eggs, and lobes of major organs like livers and lungs. And while almost no Utopian doctor will assist in ending someone’s life so that they can sell organs (and nobody is desperate enough for money to want to), it is typical in Utopia for people to sign up to auction their organs when they happen to die for other reasons, helping to pay for funeral services and other end-of-life costs, as well as providing an ample supply of other organs. As a result, Utopians are healthier and wealthier, with only the outlier communities experiencing any organ or tissue shortages.

As a last note, I think it’s worth recognizing that the same cultural and political norms that allow for more permissive body markets also support and encourage professional surrogacy, which is much more common in Utopia. The universal government protects children from being traded for profit, but there is nothing wrong with paying a woman to care for your child, including in her womb. Surrogacy arrangements almost certainly improve the overall fertility of society,17 and Utopia cherishes any opportunity to have more happy, healthy children in the world.

Prerequisites: Advertising, Opinions, Customizable Software

In my essay on advertising, I presented an argument for why Utopia has almost no advertising. Instead, Utopians must rely on word-of-mouth and opt-in systems for picking products, such as professional reviews, rating platforms, and recommender systems. I can think of no systematic ways to improve qualitative reviews, whether word-of-mouth from friends or published online by professionals, and thus I believe they are very similar in Utopia. Being holistic and idiosyncratic, it’s hard to do better than “just say what you think.”

But Utopia, like our world, also has systems for aggregating data across many consumers to provide numerical rankings and automated recommendations. And, unlike our world, these Utopian technologies aren’t weighed down by a bunch of dumb conventions and coordination problems.

Starting with…

Five-Star Reviews Suck

Let’s say that you go to a restaurant and the food they serve you is bland and unappealing, but there’s nothing particularly bad about it. You wouldn’t want to eat there again, but you’ve had worse meals. It was thoroughly mid. That’s a 3 out of 5 star experience, right?

(Or, wait… Can you give something 0 stars? If so, the middle would actually be 2.5 stars. Or maybe you can’t give 0, but you can give half stars, in which case it’s sorta 1 to 10, which would make the middle 5.5 out of 10, a.k.a. 2.75 stars. Does that mean you should randomize your response? If only there was some sort of clear cultural norm about what the lowest counting number is!)

But now let’s say you saw ahead of time that the restaurant was rated 4.8 stars! Ugh! Your experience was so much lower than that! You, very reasonably, want to press that rating down towards the middle. How can you do that most effectively? By leaving a 1 star review, of course! And similarly, if the rating is currently at 1.2, wouldn’t you be tempted to at least give them a 4 star review to help lift their mediocre-but-not-awful food out of the rating dumpster?

(Also, doesn’t having low/high expectation also tend to distort the scale? A fast food joint might get rated 5 stars if it’s clean and functional, whereas a fancy restaurant might get 2 stars if their selection of wines was lacking and their salad entrée wasn’t very flavorful. Many famous attractions are notorious for getting disappointed reviews.)

Perhaps, taking an Uber, the driver is playing a kind of music that you particularly hate, and you have this uncomfortable interaction with them about it that you think is probably mostly on you. Do you give the ride a bad rating, knowing that even a small number of bad reviews can really tank a driver’s ability to get rides and earn money?

The fix to these problems is extremely simple: show users the median rating (and ideally the full histogram) instead of the mean. When services and products are ranked according to the median, a 3 star review has exactly as much power as a 1 or 5 star review, lifting up at full strength if the review is below 3 and pulling down equally hard if it’s above. Unlike in the current system, there is no reason, in a system that uses the median, to distort your rating towards the extreme.

This also helps put outliers into context. If someone happens to love a garbage movie, their 5 star rating won’t really affect the bottom line. If you have an unpleasant experience with an otherwise good Uber driver, that low rating won’t seriously dent their record.

If the rating system is very coarse, like only having five options, showing the median won’t be very informative. Most comparably-good options will have the exact same score, and thus be indistinguishable. Now, it’s arguable whether this is even a problem in all contexts.1 The difference between 4 star dentists is probably less important than how far they are from your house, and we shouldn’t fetishize precision in noisy domains. That said, the solution (where it’s needed) is obvious: give people a more granular rating system, like -100 to +100, or a smooth slider from -1 to +1.

A system that incentivizes people to only give extreme ratings will naturally devolve into a binary “was this acceptable” check — effectively a thumbs-up or thumbs-down. Five-star rating systems are bad, but like vs dislike (or like vs no like) systems are even worse! Having no granularity encourages pandering to the lowest common denominator and making conservative, bland choices. Will the unnecessary sex scene in the back-half of your novel make anyone more likely to “like” it, or will it simply make people who already like it (or who would rate it 5 stars) even more into it, while simultaneously bringing a storm of dislikes or 1 star ratings from the mainstream?

Goodhart Ruins Everything

Regardless of the scale or mechanism, the use of numerical ratings as a primary way that we select which products to buy also creates pressure towards safe blandness2 due to the threat of a large number of people dogpiling on a product, often without actually engaging with it, and review-bombing it to oblivion because of some controversy. This most commonly happens around artwork, like books, films, shows, and especially video games, but can also show up around merchants that express the wrong political stance, or products that are affiliated with some controversial figure. Want to avoid your movie getting canceled? Better make it as unobjectionable as possible!

The flip side of this also exists. Works are pre-released to paid influencers and audiences with a known bias in order to cultivate early positive reviews. Hollywood leans into existing IP in order to take advantage of fans who go easy on characters and storylines they already like. More underhanded are bots and review farms, or even simple deals for those who leave positive reviews, which can spark endless games of cat-and-mouse with review platforms (if they actually care).

I remember the last time I moved house, I picked a service off Google with a bunch of positive reviews. They were indeed good, but I came to understand how they’d managed to get so many positive ratings when, at the end of the work, they made this huge deal out of me leaving them a positive review right then and there, with them offering to walk me through the process and pressuring me to go into detail about my experience. In a similar way, I’ve bought products on Amazon that have had big notices attached reading “CONTACT US DIRECTLY TO RESOLVE ANY ISSUES! DO NOT LEAVE A NEGATIVE REVIEW!” On one level it’s nice to have good service, but these are signs that our systems are falling victim to Goodhart’s curse — producing interactions that are good at generating positive reviews, sometimes at the expense of the product or service.

These kinds of dynamics go on and on. Creators set up networks of mutual-promotion, trying to game the algorithm. Guests staying at Airbnbs leave good reviews because they know their hosts are, in turn, reviewing them.3 People leave positive reviews out of a mistaken urge to be nice.4 Youths leave extreme ratings on works of mediocre quality out of a desire to express their identity. The distortions can seem so omnipresent that one might despair of all efforts to escape the incentives, even on the margin.

One Number???

To round out our tour of the ways that our culture is dropping the ball, let’s turn our attention to what is probably the worst part of rating systems: compressing everything into a single number.

Taste is not the same as quality. I might be able to tell that a couch has good craftsmanship and materials while simultaneously wanting the style to stay far, far away from my living room. 2001: A Space Odyssey was a masterpiece, but is painfully slow and boring from the perspective of most modern viewers. Store brand soda doesn’t have exactly the same flavor profile as a big brand, but it’s certainly cheaper. (Does that mean it’s higher quality? The same quality? Is preferring Dr. Pepper over Mr. Backwash simply matter of taste?) When we ask someone to put a numerical rating on something, what are we even doing?

The sophisticate attempts to break things down into sub-properties, hoping that by doing so, they can be more objective and let others decide what’s important. But outside of a few obvious dimensions like price, these decompositions strike me as a poor approximation of good-old qualitative reviews. What is the “replayability” of a game, really? Is the replayability of chess higher or lower than Minecraft? Is the acting in a film bad, or is it the director being bad at taking advantage of skilled actors? Or is the wooden acting part of the point, like in Napoleon Dynamite? The fractal character of reality resists being decomposed into a small set of numbers, and attempts to do so can be corrosive. There is information in the gestalt.

What consumers need are scores that are decision-relevant, and which track the specific aspects that they care about — no more. What do people with similar tastes like? What have they already bought/experienced? Do they have any idiosyncratic preferences, like a phobia or particular bit of childhood nostalgia? All of this is a hard problem that requires intelligence and effort. Even customized recommendation software, like on Amazon and YouTube, cannot account for things like what mood you’re currently in, whether you want to be challenged vs entertained, or whether you’re with other people.

Making the problem even harder, things change over time. Preferences drift. Stores come under new management. The zeitgeist unfolds. How can any numerical system hope to track so many different variables and combine them in the right ways?

Prediction Markets?

Long-time readers will, at this point, be anticipating the obvious thing to try: prediction markets. Where a numerical rating can be casually tossed out there with little effort or regard for the truth, a bet has teeth. Failing to see quality or not appreciating what other people like becomes expensive. The involvement of money also solves a major problem that we haven’t even discussed: correctly rewarding people for doing the work of leaving a high-quality review. With enough money on the table, those with a discerning eye will naturally be incentivized to lend their expertise towards picking the best of the bunch. In theory, prediction markets can even solve the problem of personalized tastes and moods. Simply make a prediction market on what you (or someone you know) will enjoy, and tell the world what you’re into.

But prediction markets are not a panacea. Many of the problems we’ve been discussing, such as the extrinsic pressures, aren’t helped by the use of markets, and new problems can show up on top. For a market to be a good signal for the truth, it needs to be “thick.” There needs to be enough money at stake for “sharks” — domain experts with skill in playing the marketplace — to spend precious time and attention holding the probability in the right place.

Many of the use-cases for recommendation systems are naturally “thin.” If I’m looking for what restaurant to go to for lunch, domain experts won’t have the time to notice it and weigh in according to my tastes. There’s a reason that the Manifold markets I linked to in the previous paragraph are about books, rather than music or restaurants. The generally low-information environment and the presence of insiders also discourages anyone from betting who doesn’t already know you well, contributing to a streetlight effect that is even worse than the one we see on today’s rating sites.

To attract domain experts, one must put up significant money, possibly in the form of subsidies. But with more money on the line, extrinsic pressure to distort the market intensifies. If I’m putting $100 down to create a market on whether I’m going to like my mattress, I might be awfully tempted to say I did/didn’t like it just to recoup my money by betting on the unlikely outcome (or having my friend bet for me). If something is controversial enough to attract a high volume of trading, people might try to sway the outcome directly, whether through review-bombing… or more extreme measures.

Utopian Reviews

The solutions to these problems start with a fundamental restructuring of software, privacy, and data ownership.

In our world, companies are assumed to be the ones that track what their customers watch, buy, and like. If you use a credit card when you go to the movie theater, the credit company will track that you bought tickets. If you use a rewards program, the theater company will track which film you see. This data is then sold to third parties, usually for advertising purposes. By contrast, Utopian services use better forms of identification and payment, such that companies not only lack an incentive to surveil, due to the de-facto ban on advertising, but they are actively incapable of doing so. (It’s also semi-illegal.)

But this doesn’t mean that nobody has access to that data! The typical Utopian uses streamlined (and personalized) data-management software that holds onto a timestamped record of everything they’ve seen, listened to, bought, everywhere they’ve visited, and everyone they’ve talked to.5 By owning that data themselves, and having standardized protocols for sharing it, they can provide limited access to that information to others, such as by curating end-of-year lists of favorite music, or quantified-self-style findings about how their taste in books has changed over the years.

Data vaults also serve as a foundational platform for keeping notes on absolutely anything. Want to make plans for Wednesday? Want to remember your friend’s phone number? Annotate vacation photos with who’s in them? Utopian note software has you covered, without needing to trust a third party.6 When relevant, these notes can be published as online reviews. The same cryptographic authentication that underpins the identity and payment systems in Utopia allows for reviewers to verify that they’ve actually purchased the product that they’re reviewing.

Consumers can also annotate their notes with a set of discrete reactions, called “tags.” Thanks to the customizable, decentralized nature of this sort of software, there are a whole host of bespoke tags, both public and private that are used by various communities. Here are my guesses as to the default-public ones that are mainstream:

• High Quality — The tagger thought that this was particularly well made, and reflects skill on the part of the creator (regardless of whether they liked it).

• Favorite — The tagger really, really liked this thing. Utopians often like to share their lists of favorite things across various domains as a way of getting to know each other. By having a distinct notion of High Quality, Utopians are less likely to get into fights about whether other people’s favorites are worthy of the attention.

• Classic — The tagger thought this is culturally foundational. It’s generally understood that Classics are important for most people to be at least a little familiar with, for the sake of having shared cultural references, even if they’re not as engaging or fun as other works.

• Fresh and Exciting — The tagger liked this in a way that was connected to its novelty or interestingness. There’s a general heuristic in Utopia that it’s wise to, in the heat of the moment, tag things Fresh and Exciting, then only after many days (or even years) have passed, reflect and promote some of them to High Quality or Favorites, and then only after decades of reflection, promote some of those to the status of Classics.

• Warning — The tagger thought that this thing was likely to be particularly difficult for some people to deal with. The exact nature of why something is dangerous is usually left to qualitative reviews.

• Low Quality — The opposite of High Quality. Things that are Low Quality might still be worth it, if they’re cheap, and they might even be fun, or “so bad it’s good.”

Finally, while it’s not exactly a tag, there is also one bit of similar machinery that’s nearly ubiquitous in many kinds of professional interactions: gratitude tokens. In Utopia, when you benefit from a service/customer, it is customary for both parties to give a token of gratitude, usually in the form of a simple digital message.7 While gratitude tokens are similar to giving a good review or hitting a “like” button (or *shudder* giving a tip), they are explicitly meant to be distinct from an endorsement of the service or product. Rather, they are intended to be a pure, gracious acknowledgement that someone else did something to improve your life.8

It’s hard to measure gratitude, since (by default) it’s private between the two parties. Still, a movie-maker might notice that one of their films collected far more gratitude than another.9 Far easier is tracking tags, and some Utopians make use of algorithms similar to PageRank to figure out what shows have, for example, the highest weighted-fraction of Fresh and Exciting tags in their community. But in general, people in Utopia recognize that trying to make decisions according to a single, dumb aggregate number is a somewhat foolish strategy. Instead, they base their picks on a gestalt of what they hear from friends and professionals, and they make use of advanced recommender systems.

Utopian Recommender Systems

Like our world, some services in Utopia have built-in recommender systems. Internet radio, for instance, might craft an instant playlist based on what’s known about the listener’s tastes. Thanks to more universal standards, it’s easier to port data from one service to another in Utopia. Favorites are stored in the user’s data vault, after all.

But the biggest and most popular recommender system in Utopia, which I’ll call “Avu,”10 isn’t really a single piece of software at all; it’s a prediction market platform that only bots can use.

Suppose that one evening, you and your date are in an unfamiliar part of town, trying to decide where to have dinner. You open Avu, and ask for restaurant recommendations, noting that you’re in the mood for live music. Your chosen assistant creates a handful of prediction markets that ask things like, conditional on eating at [place], whether both of you will unambiguously enjoy the food, or whether either of you will tag the restaurant Fresh and Exciting. The creation of these markets then spurs various third-party AIs to evaluate whether to invest energy in betting. If they do, fresh AI agents are spun up, which immediately get to work in reading your notes and scanning other databases, both public and private, looking at menus, trends, weather, and anything else they think might be helpful. The bots send your date a push notification asking them to give access to their notes to get better answers.

AIs are an excellent complement to prediction markets. In late 2025, they were already out-competing median forecasters, and are getting nearly as good as the best human experts. Where a couple-dollar bounty would be far too small to pay for a human expert’s time and attention, an AI model that can, in expectation, earn a few cents of profit might make the company hosting that model filthy rich. AI is fast, capable of reading thousands of pages of notes and synthesizing information from dozens of sources in mere seconds. And, perhaps most importantly, AI can be kept sandboxed from the rest of the world, accessing private information without risk of it leaking.11 Trained to operate in the enclosed environment, these agents are unlikely12 to do anything beyond try to make a profit by accurately predicting what you’ll like.

Avu gets back to you; the bots have spoken. An ethnic bar pops up as the front-runner on Fresh and Exciting, with a comparably high probability of enjoyment when compared to a chain restaurant that both you and your date have independently enjoyed. You click a button, paying $0.2813 to get an explanation from the AI that was most confident in it being a good choice.14 It explains that the bar has food that it knows you and your date both enjoy, and gives recommendations. It goes on to say that it’s not too crowded right now, based on the bar’s own data feed, that the band that’s playing there tonight recently got a rave review from a local critic, and that they play a genre of music that your date recently lamented not getting to dance to frequently enough.

Several hours later, Avu sends a reminder to share your experience, wherever you ended up actually eating. There’s even a small monetary incentive attached — credits that you can use to buy your next recommendation. This reward scales based on market volume, so when bots disagree strongly and trade volume spikes, people can be heavily compensated for giving their honest notes. (This can also nudge people to go outside of their normal groove to try something weird.) Once your notes are sent over, an impartial judge-bot is then prompted to use those qualitative notes to resolve the relevant markets.

Utopian businesses occasionally try to trick Avu by doing sneaky things like under-reporting how busy they are, paying influencers, or pressuring people to write nice reviews. But unlike our world, where consumers don’t have the expertise or energy to spend time sniffing out these kinds of bad behaviors, Utopia’s Avu companies can get a repeatable edge in the market by hiring investigators with a track record of looking into these things. A judge that has been (verifiably) informed by one of the traders that the company is pressuring buyers to write glowing reviews is less likely to treat a glowing review as a strong signal that the buyer was satisfied.

Because notes are usually private, users can write freely about their honest experience. The more information someone puts in, including by being disagreeable and panning things they genuinely don’t like, the better the bots will be at using that as a signal to select high-value opportunities in the future. Many people even write down their pre-registered expectations as a way of providing better data about whether something was surprisingly good, and getting away from the various biases of expectation.

Storefronts (and areas of Utopia where advertising is allowed) take advantage of Avu by creating “typical shopper” stand-in characters and highlight merchandise that the bots think that these default buyers will prefer. Celebrities and famous critics also sometimes sign deals to use their Avu data for this kind of promotion (which Utopian elites find a bit gauche to attend to instead of querying Avu directly). Thus, instead of seeing that a jacket has “4.3 stars,” a Utopian shopper might see that it’s the jacket that Avu thought would be most likely to be described as fashionable and High Quality by the star of a recent action movie.

Still, despite the superiority of Avu, there is a common wisdom in Utopia that it’s a mistake to let high numbers dominate one’s thinking too heavily. To learn and grow, sometimes it’s important to embrace spontaneity and be willing to try things that aren’t guaranteed to please.

Prerequisites: Having read Spacetime Distances will probably help, but isn’t necessary. I go into a lot of physics, especially in the later footnotes, and that might be more than many readers can comfortably handle, but I’m trying hard to make it as accessible to a general audience without glossing over the important bits.

Learning is the process of colliding with reality so that a model of the world forms in the mind. This world-model, which is essentially all you have conscious awareness of, is about the world, but it’s important to notice that it is not actually the world. The map is not the territory in many different ways.

• Maps contain errors and omissions. The map can have a forest where the territory has a lake. When a map corresponds closely to reality, it is more “true.”

• Maps fold up and fit in backpacks. This makes them convenient to work with, but also creates distortions that aren’t exactly errors, but can definitely still bias things if you don’t account for them.

• Maps are naturally covered in labels, lines, and other notes. These notes are usually about real things, but they aren’t, themselves, literally in the territory.

In physics, the core principle of relativity is so simple and obvious that we often take it for granted: to compare maps, we must ignore the notes and try to line up the parts that are objectively true. The actions that we take to get the maps to line up should lead to both models having the same measurements — sym metries.1

A symmetry is a way we can change the superficial characteristics of our models without making them any less true. The relative distances between cities are objective facts, but the choice to put north at the top of the map is arbitrary. We are free to translate and rotate “the whole world” (which really just means moving our perspective) without any real effect.

But symmetries imply something even more powerful than the ability to see the whole world from a new angle. If two map-makers make different choices about how to represent the world, it indicates that there is a dimension along which parts of the world (that are otherwise very similar) can differ. The ability to move the lines of our Cartesian grid is a hint that the position of individual objects is variable — if you see an apple in one location, you can see an apple in another location. The ability to rotate our coordinates implies that objects exist at a variety of angles and orientations. The fact that we can flip all the coordinates and preserve the relative distances between things means that if we see a left-handed glove, we should expect a right-handed glove to also be possible.

Right?

Clockworld

Let’s set aside our universe for a bit and consider a two-dimensional world inhabited by various fundamental objects, like triangles and spirals and so on. Just as in our world, we can imagine translating, rotating, and reflecting the space, without really changing the way these objects interact. As long as the positions, velocities, and angles are the same relative to each other, things will play out in the same way.

Let’s now introduce an arrow-shaped object that we’ll call a “clock” because of a special property: if a clock object spins anticlockwise, it will quickly delete itself. Thus a box full of clocks that are spinning in all sorts of ways will quickly become a box full of clocks that happen to be spinning clockwise, because any that are spinning the other way will disappear.

In physics lingo we say that the self-deleting rule for clocks “breaks parity.”2 Because the orientation of the clock matters for what actually happens, it means that we can no longer reflect the universe and consider that perspective to be equally valid under the same laws of physics.

When parity seems to be broken, there are a couple options available to us. One is to shrug and conclude that “reality picks a side” — that there’s something fundamentally real about the orientation of things, beyond their relationship to each other. Another option is to guess that the supposed “violation” of symmetry is actually a clue that you missed something. For example, what if the supposedly “fundamental” clocks are actually a combination of two parts: a big, obvious clock-hand, and a tiny, internal clock hand that is also turning clockwise?

This change lets us frame the clock rule as about the relative properties of the parts: “If the big hand is turning opposite the small hand, the clock destroys itself.” Or, if we don’t want to think of the internal hand as a separate object for some reason, we could alternately say that the clock has a “clockwise flavor” and the rule is “If the clock spins opposite its flavor, it breaks.” Regardless, the key here is that when we reflect, we must also flip each clock’s “clockwise only” rule to become “anticlockwise only” to ensure the mirror universe has the same outcomes.

Attaching the rule to the clocks themselves, rather than the universe as a whole, is aesthetically satisfying in addition to keeping the physics consistent. The universe seems to respect the obvious symmetry in all other cases, so it’s much cleaner to find a story for the behavior of the edge case that is genuinely about the edge case, rather than letting it shatter the core principle.

But symmetry isn’t only about global changes. If it’s possible to flip the orientation of all the clocks — whether via internal hand or flavor or whatever — couldn’t we flip only some clocks? The restoration of symmetry predicts there should be anti-clocks hanging around in the same world that the clocks inhabit.

The story of clockworld is fictional, so we can freely decide that anti-clocks do or do not exist alongside normal clocks. But it turns out that there’s a remarkably similar phenomenon in the real world, and it probably isn’t the one you’re expecting…

A Quick Note On Handedness

In our universe, almost everything has “angular momentum” — a quantity of spinning, analogous to the way objects can spin in the 2D universe (but, unfortunately, with a bunch more complexity that’s mostly irrelevant to our current topic). For example, each atom in a hunk of metal is composed of a cloud of electrons surrounding a tiny, central nucleus. The nucleus, in turn, is made of a dense swarm of protons and neutrons, which are themselves made of a cloud of quarks.3 Each of these objects — whether it’s an atom, electron, nucleus, proton, neutron, or quark — has an angular momentum.

By convention, we use “the right-hand rule” to relate rotation to (linear) direction.4 If you curl up the fingers on your right hand in the (angular) direction of rotation, we say the object’s angular momentum is pointing in the direction that your thumb goes if you stick it out like a hitchhiker.

Notice that if we could actually see a ghostly right-hand around some spinning object, and that hand always followed the right-hand rule as a law of nature, this law would violate parity. If we place a mirror in front of the object (parallel to the plane of its rotation), the version in the mirror will still, from our perspective, be spinning the same way. Thus the ghostly hand won’t flip. But the ghostly hand, by the rules of basic reflection must have flipped. Contradiction!

The right-hand rule is just a convention, however, so as long as we only use it to relate the rotational dynamics of different things in the same universe, we’re fine. For example, a rotating charged object (or a spinning electric current, like in a solenoid) will produce a magnetic field5 in the surrounding space. This magnetic field is naturally about spinning, so we can use the right hand rule to pick out which way is “north.” If we place a hunk of metal in the field, the (electrically charged) quarks in the metal will be pressured to move in ways that makes the angular momentum of their nuclei either match the direction or point in the opposite direction of the field. But despite the language of “direction” this is just spinning affecting spinning, akin to an assemblage of gears.

The Wu Experiment

At the temperatures we normally encounter, the kinetic energy of the particles in a hunk of metal will usually mean their rotational directions will be super random, even if there’s pressure from a magnetic field. But if the metal is sufficiently cold and the magnetic field is strong enough, more and more of the particles in a chunk of metal can be made to consistently spin in a desired direction.

Because the protons and neutrons in a nucleus are bound tightly together, it usually makes sense to consider each nucleus as a whole. And if the metal nuclei happen to have a positive gyromagnetic ratio,6 they will spin clockwise from the perspective of someone looking from the south end of the magnetic field towards the north end. In the language of the right-hand rule, we say they’re pointing north.

Again, the right-hand rule is a human convention. Everything shown in that picture is actually reflectively symmetric, including the magnetic field.

If the metal in question is also radioactive (eg Cobalt-60), neutrons can spontaneously decay into protons by having one of their “down” quarks change into an “up” quark, spitting out an electron and a neutrino in the process.7 To contrast with the strong force that holds the nucleus together, the mechanism that produces this radioactive decay is considered part of “the weak force.”

If you'd asked any physicist in the middle of the 20th century whether the electrons radiating from the metal would be biased towards the north vs south end of the experimental setup, they would have rejected the notion. How could there be? Everything involved is symmetric!

Chien-Shiung Wu, in 1956, was the first person to actually check. And as it turns out, nearly all of the electrons go south. The standard explanation is that even though all the objects are symmetric, the weak force itself is not. The right-hand rule (or rather, the left-hand, since the electrons go south) was declared to be a real law of nature that violates parity.

Nobel laureate Wolfgang Pauli (who almost bet a large amount of money on the outcome) was one of the many scientists to be not only surprised, but deeply unsettled. Just before the experiment, he wrote: “I cannot believe that God is a weak left-hander.” And later: “Now after the first shock is over, I begin to collect myself. Yes, it was very dramatic.”

After a failed attempt to rescue parity by trying to unite it with the matter/antimatter split, much of the physics community shrugged and declared parity to be dead.8 But this is not the only option. If we want to rescue symmetry, we must turn to an even more bizarre story. Thankfully, most of the weirdness is already included in the standard model, and well validated by experiment.

The Fabric of the Universe

As I discussed in my essay on smoothness, we should be skeptical, on principle, of any story of matter that treats it as made of discrete particles. Indeed, a more real picture is that what we usually think of as “stuff” is more akin to ripples moving along a piece of fabric. Just as light can be seen as waves in the electromagnetic field, electrons and quarks can be seen as waves in their respective “Dirac fields.” In this picture of reality, “empty space” — vacuum — is merely a part of reality that doesn’t have ripples. But, importantly, it still has fabric.

Not all possible vacua are identical. One of the most celebrated predictions in the history of physics was the 2012 observation of the Higgs boson — the particle (ie quantized excitation) of the Higgs field, first proposed in 1964. Though Higgs bosons are extremely rare, the character of the Higgs field itself constantly influences the other fields — the vacuum itself can be seen as having an omnipresent structure. And importantly, this structure has a kind of directionality built into it, as though the weave of the fabric has an orientation.

In the very early stages of the universe, the Higgs field is theorized to have been more symmetrical — lacking this intrinsic orientation, but having a higher intrinsic energy. But as the universe cooled and expanded, the vacua in various parts of reality settled into more stable configurations in a process called “spontaneous symmetry breaking.”

We can see a similar phenomenon at a more human scale when hot metal cools into crystals. In its heated state, the orientation of atoms in the metal is chaotic and without any directional bias. Then, as the metal cools, atoms nest together in a lattice of repeating patterns. When crystals in different parts of the metal have different orientations, a boundary forms between them, where the orientation suddenly shifts. (Blacksmiths repeatedly heat and cool metals to manage these boundaries, as they can lead to brittleness.)

While the asymmetry in the Higgs field is more complex than a simple direction in space, and usually isn’t used to explain the asymmetry of the weak force… might there, nonetheless, be a way in which our particular vacuum secretly forces a handedness to things? If so, we could easily explain the supposed parity violation of the Wu experiment as an oversight — we forgot to flip the vacuum!

Left-Right Symmetric Models

In the standard model of physics, the electron field can be viewed as two distinct (sub)fields, which we label as either having a left or right “chirality.” This use of “left” and “right” comes from the way these components relate to the right-hand rule for angular momentum. As usual, reality is more complex than a summary can capture, but at a very high level, right-chiral particles want to travel in the (right-hand rule) direction of their spin, and left-chiral particles want to go the opposite way. I'll relegate a more complete explanation to the footnotes.9

Regardless, it’s important to remember that chirality is about the existence of multiple types of things, rather than about how things are turning, per se. To reduce the chance of confusing right-handed spin particles with right-handed chirality, I’ll use the individual letters “L” and “R” to refer to chirality — the two kinds of matter.

Wu’s experiment demonstrated that nuclear decay produces electrons which tend to move opposite the direction of their spin. Does this mean they’re L chiral? In truth, all electrons are a mixture (“quantum superposition”) of L and R states, with energy simultaneously flowing back and forth between their two chiral fields via the Higgs field. (It’s this flowing energy that gives electrons their mass.) But particles, particularly high-speed particles like the emitted electrons in Wu’s experiment, can still be disproportionately one chirality or the other.

In the standard story, quarks, too, have L and R components. But neutrinos — the only other form of matter,10 and importantly, one that basically only interacts via the weak force — don’t. We’ve never seen a neutrino with any notable R component. In theory that means neutrinos should be massless, but we know that’s not the case from other experiments. It’s an open question both where their masses come from, and why they’re so light — the mass of the neutrino is over a million times lighter than the electron.

When we observe the weak force, we only see it coupling to the L side of matter, and in the standard model this asymmetry is baked in as an assumption. It’s also broadly agreed that the Higgs field symmetry broke in a particular way, and that this gave rise to our particular vacuum, which in turn determines how much energy it takes to create a particle that carries the weak-force (“W and Z bosons”). But suppose that these two facts are connected, and there’s a parallel version of the weak force that would operate on R matter, except the vacuum demands extremely high amounts of energy in order to push energy through it — the weave of the universal fabric is so stiff as to effectively eliminate the possibility of a ripple.

In a left-right symmetric model, this is exactly what is supposed. Nature set up initial conditions that were nice and symmetric, and then something broke that symmetry such that the L and R versions of the weak force came apart. Very nicely, because the asymmetry comes from the Higgs field,11 and the Higgs mediates L-R interaction to give things mass, the R neutrinos that we are forced to add to our picture of reality in order to keep things symmetric are both too massive to show up (in our vacuum) and simultaneously make L neutrinos extremely light but still give them some mass.12 This trick is called “the seesaw mechanism” and is so elegant that adding heavy R neutrinos is a common move when extending the standard model. Having a broader LR symmetry simply means you get them from underlying principles.

LR symmetry also gives a few other nice things13 in the same sort of way, and is a very common feature in so-called “grand unified theories” of the cosmos (including string theories). The biggest reason14 it hasn’t been accepted as a canonical extension of the standard model is that we don’t have any direct evidence — it’s all theoretical speculation.

Utopian Symmetry Breaking Resilience

Utopia is not afraid of theory, even when it’s speculative. Instead, people in Utopia tend to be far more averse to being unprincipled, whether in their social structure, personal life, or scientific theory.

A big part of the way that ideas are judged in Utopia is according to their simplicity and elegance, and symmetries are a primary source of natural elegance. With each symmetry, we can ignore some of what’s on on our maps as irrelevant detail. A circle is more symmetric than a square, and it’s not a coincidence that a circle can be specified with less information — that symmetry is a direct reflection of its simplicity. On priors we should expect the deep structure of the universe to be highly symmetric.

This is why the notion of a parity violation should be upsetting. The world is so incredibly close to looking like we could put it in the mirror without breaking anything. Parity is necessary to allow for time to be reflected, and thus for allowing a bi-directional deterministic view of the universe. This foundation is so important, that Utopians are more than happy to adopt more speculative models, rather than abide the blight of arbitrary directionality.

And so, my guess is that the Utopian response to the Wu experiment is probably similar to their response to the double-slit experiment and other indicators of quantum weirdness. They certainly don’t ignore the evidence, but they also don’t let it totally shatter the traditional ontology.15 Instead, Utopians look for ways to accommodate the new data using elegant principles, and are more likely to adopt beliefs like the many worlds interpretation or some kind LR symmetry.

And ironically, in the end, hunting for elegant methods of preserving a simple worldview is an epistemic move that has a well-proven track record. From Copernicus to Darwin, scientific revolutions have often come from a place of trying to find clean and simple stories that explain the seeming complexity of the world. When symmetry and elegance are threatened by an unexpected result, Utopia rises to the challenge with a hunt for even deeper symmetries, and the better physical models that emerge in the process.

Prerequisites: Part 1, Adjudication, section on Utopia builds on many important ideas from Basic Income, Freethought, and more.

Prisons are a nightmare.

Outside of some exceptions, like Bastøy, prisoners around the world are subject to daily indignities,1 including:

• Sweltering heat

• Freezing cold

• Intense boredom

• Harsh smells

• Starvation, malnourishment, and awful food

• Sleep deprivation

• Brutally hard and uncomfortable environments

• Torturous solitary confinement

• Lack of time outdoors or even access to windows

• Being tied or chained down for hours or days

• Forced contact with filth and disease

• Forced contact with awful people

• Lack of personal space

• Lack of privacy

• Lack of ability to really work or study

• Forced labor for little or no money (a.k.a. slavery)

• Lack of ability to make choices

• Isolation from friends and family

• Lack of proper medical care

• Exposure to hard drugs

• Being mocked and bullied by inmates and guards

• Being sexually assaulted and raped by inmates and guards

• Being beaten2 by inmates and guards

Note that all of those hyperlinks go to cases in American and British prisons. Institutions in less-wealthy places are often much worse. It should be no wonder that the suicide rate in prisons and jails is many3 times higher than it is in the broader world.

In full view of all this, it’s no wonder that reformers nearly everywhere seek to emulate the more humane Nordic model and invest in restoring dignity.

But from my perspective, one of the worst parts of the situation is that the conflict between high spending and inhuman cruelty is a false dichotomy. There are ways to simultaneously restore dignity, reduce recidivism, and keep costs low. This essay is about how.

Universal Rights

Before getting to prison reform per se, I want to take a detour through the notion of rights. A right is a normative principle that specifies what duties4 (actions that are either forbidden or mandatory) some party has towards each person in some class (the holders of the right). For instance, a “right to free speech” might stipulate that the state has a duty not to prevent a citizen from expressing their ideas to other people. A “right to clean water” might stipulate that restaurant owners have the duty to serve tap water to customers without charging money.

Regardless of whether rights are natural5 or merely a legal construct, they are an undeniably foundational part of the social order. Laws that are based on solid principles will be clear and compelling, where laws that don’t have a principled backing will inevitably fall apart and turn into a jumbled mess as soon as people start picking at the edge-cases. Of course, one can always construct a “principle” to justify something. I could argue that I have a right to your sandwich. But principles are not created equal, and just as unprincipled laws/societies are naturally at risk of fraying into legalistic confusion, those that are built on weak principles are at risk of sparking rebellion due to lack of fairness or being crushed by endless red-tape and burdensome obligations.

Good rights are those that:

1. Focus on solving some kind of coordination problem, such as preventing wasteful competition or establishing common knowledge/standards/expectations.

2. Assign duties narrowly, ensuring that everyone understands who is responsible.

3. Help the broadest and most obvious set of people, maintaining fairness and avoiding membership tests.

I see this last meta-principle as particularly important. Historically, improving universality has perhaps been the biggest source of moral progress, and I bet that insofar as the future is good, a lot of that goodness stems from extending rights to be more universal.

I’ve written about universality in several earlier essays. In Personhood, I argued against a notion of “human rights” and instead for “people’s rights,” allowing for the inclusion of non-human people. In my sequence on Basic Income, I argued that by providing welfare in a truly universal way, such that all people get the same support, we can reduce unfairness, inefficiency, and stigma. In Futarchy, I presented a system of government where children (and the mentally disabled) have just as much influence over the values of society as anyone else, by having their parents/guardians vote as representatives. Here, I want to argue that the pattern of depriving prisoners of rights is yet another instance where our society is dropping the ball, morally.

Consider the very common practice of excluding felons from the set of people with a right to vote. Justifications for this tend to come in three flavors:

• “Criminals have violated the social contract and are thus not entitled to the benefits of society (perhaps additionally serving as retribution or deterrence).”

  • Counterpoint: This argument is too broad. It can be used to justify arbitrary abuse of criminals, regardless of the nature or severity of their crime.

• “Criminals have demonstrated a lack of wisdom and a divergence of values from the rest of society; we don’t want them to mess up collective decision making.”

  • Counterpoint: We emphatically do not deny democratic participation to members of society that are unwise or have different values. Ensuring everyone has a voice is the whole point of democracy.

• “This is the way it’s always been, and other people do it, so it’s clearly okay.”

  • Counterpoint: This argument denies moral progress and has an extremely bad track record, in the context of history.

Furthermore…

• Making sure all citizens have the right to vote, no matter what, feels like the moral high-ground on the vibes-level. Aesthetics and vibes matter. If you don’t encourage virtue signaling, what you get is vice signaling.6

• By denying some people rights, based on their personal history, you introduce inefficiencies into social systems. Admittedly, in the case of voting this is less of an issue because we already exclude large classes of people (eg foreigners), but in some cases it can introduce significant costs.

• Certain characteristics correlate with being a felon, and by stripping felons of their rights, this inflicts splash damage on those communities in ways that at least runs the risk of being seen as unfair. (And is plausibly just straight-up unfair.)

• Allowing the government to remove political power from certain people is a dangerous game, and paves the way to corruption and tyranny.

• Denying felons the right to vote amplifies the narrative that they are outsiders and enemies of society. This is contrary to the goal of rehabilitation and increases the risk of recidivism.

This final point can be generalized to a broader principle: normalization. The more we treat prisoners as a distinct group of people and make their lives extremely abnormal, the less likely they are to re-integrate into healthy social roles upon being released.

Consider the way that most prisons heavily restrict the agency of convicts, dictating where they must be at every hour of the day, what food they must eat, and so on. After years of being forced to follow a strict routine, people become institutionalized, and can develop persistent issues with willpower and impulse control after being released.

I am not arguing that money should be spent to give prisoners a diverse curriculum of opportunities to choose from. Having a menu of options for classes is also not even what normal life is like. (Arguably, young people leaving school often have very similar signs of institutionalization!) Rather, I am arguing that there is effort (and thus money) currently being spent controlling prisoners’ lives and denying them the same rights as others, where the wiser option would be a smarter, more hands-off approach that preserves their dignity.

Safety and Trust

Unfortunately, the people who were dangerous enough to commit crimes and land in prison tend to be, well, dangerous. The biggest issue with “be more hands-off” is that prisoners might harm each other (and themselves). Omnipresent cameras and guards are both costly and hurt normalization, but with fewer watching eyes come more opportunities for abuse.

This security vacuum is most naturally filled by the prisoners themselves, often in the form of gangs. To stay safe, each inmate in a more laissez-faire prison must often submit to one or more gang leaders in various ways. And while this can provide something like security, the presence of multiple gangs in the same prison can mean the spontaneous eruption of large-scale violence.

Perhaps worse, prisons where gangs fill the power vacuum become natural hotbeds of organized crime, corrupting both inmates and guards, and increasing criminal activity beyond the prison walls. As drugs, weapons, phones, and other contraband becomes more common, violence can become more lethal, and it can be harder for inmates to turn over a new leaf.

This approach to imprisonment (sometimes called “warehousing”) can even end up more expensive in the long run. In the United States, for example, prisons that are more lax can end up getting sued for violating the 8th amendment’s prohibition on cruel and unusual punishment. In addition to individual payouts in the millions of dollars per case, judges can also then compel prisons to reform to a higher standard of care, costing billions in restructuring.

Still, might there be some promise here? What if instead of trying to suppress gangs and the natural desire of inmates to create safety and order themselves, prisons encouraged inmates to be their own wardens? This idea is less radical than it might sound, and has been done “successfully” many times all around the world. These systems are almost always much less expensive to the jailers, allowing shockingly high ratios of prisoners to guards, but they also typically come with high levels of abuse.

Some of the more prominent examples include:

• The Funktionshäftlinge system in Nazi Germany's concentration camps.

  • There, kapos — Nazi collaborators — would supervise and discipline other prisoners in exchange for special privileges. In many cases these functionaries felt guilt and used their meager powers to protect their friends. In other cases they tortured and killed their fellow inmates. Regardless, “kapo” remains one of the harshest insults in some Jewish communities to this day.

• The convict officers of the British Empire, particularly in the penal colonies of Australia, Singapore, and the Andaman Islands.

  • Surrounded by the natural walls of the ocean (and inhospitable desert), prisoners were allowed more freedom to move about, and were tasked with farming, construction, and other labor.7 To keep the huge prison populace in line, obedient prisoners could earn “marks” that allowed them to rise in a quasi-military hierarchy of policing and discipline.8

• The trusty system9 of the American South.10

  • Only fully abolished in the 1980s, the trusty system was a way to profit from the cheap labor after convict leasing — itself an attempt to regain cheap labor after the abolition of slavery — became politically unpopular. At the top of the hierarchy were “trusty shooters,” armed with rifles and often selected for being murderers or otherwise violent enough to intimidate those working under their muzzle. While those convicted of murder often faced life in prison, for a trusty shooter, killing another prisoner could mean a cash bonus or even being rewarded with freedom. Those who were given power frequently abused it, sometimes torturing or raping the less privileged inmates.

While these examples are undeniably dark, I think it's worth recognizing that the abuse in these contexts stems more fundamentally from the way that these societies had little interest in the dignity or welfare of the prisoners. Indeed, places with different systems but with a similar level of apathy tend to also be hellholes, whether due to gang activities or mistreatment by guards.

And there are also positive examples of prisoners taking part in the hierarchy of the prison administration. In the early-mid-1800s, a man named Georg Michael Obermaier became a prison governor in Bavaria. Where the previous prison system was harsh and controlling, he removed chains and dismissed nearly all of the guards, appointing prisoners to be in charge of productive workshops and to serve as monitors to ensure good-conduct. Prisoners were treated with dignity and respect, and given progressively more freedom as they demonstrated trustworthiness, leading to lower levels of violence and bad behavior, as well as recidivism rates comparable to modern Norway. While some of Obermaier’s success was likely thanks to his charisma, his reformations were undeniably good at simultaneously helping prisoners and cutting costs, and promoted the concepts of parole and normalization.

Even more noteworthy is the Association for the Protection and Assistance of the Convicted (APAC) in Brazil. Founded in 1972, APAC now manages about 70 prisons with an average of about 100 prisoners each. And between all of them, there are zero guards on payroll. The prisoners, some serving life sentences for heinous crimes, are not only in charge of maintaining good conduct, but are also given the keys to their own front gate and trusted to not escape. Thanks to this, the labor of the prisoners, and some volunteer work, prisoners under APAC only cost about $2300 per year — less than half of Brazil’s national average of ~$5000 pppy.

This system largely works thanks to carefully selecting among many applicants, eager to escape Brazil’s awful prisons, and the recognition that escaping is most likely a path back towards serving more time in a nastier place. But it is also thanks to a self-perpetuating culture of dignity. When a prisoner first comes to an APAC prison, they are not immediately trusted, and are instead assigned to a strict schedule involving a lot of work. It is through this work that the prisoners get to know the new member, and learn whether he or she can be trusted with more responsibility and autonomy, such as dangerous tools, time unsupervised, or ability to leave on authorized trips. Those who show that they understand and have internalized the culture of dignity are allowed more freedom, and eventually go on to mentor and supervise newer inmates and have access to the keys.

Despite some claims to the contrary, these prisons have a significant number of escapes. In this master’s dissertation that investigates one facility’s record over 20 years, we can see 84 true escapes and 312 cases where prisoners who were released on temporary trips failed to return. If we focus entirely on escapes, which came from the ~90 inmates (at any given time) who had not yet demonstrated trustworthiness, that’s an annual escape rate of nearly 5% — close to a hundred times the US average. If we include the ~95 more trusted inmates that fail to return as escapes,11 the rate jumps to over 10%. Still, the vast majority of the prisoners choose to stay, which is a noteworthy accomplishment given the total absence of guards. Part of what keeps inmates from bailing is the fact that escapes result in loss of privileges for everyone on duty, meaning there is a sense of personal betrayal whenever someone leaves.

APAC’s recidivism numbers are better, though like the breathless reporting for Bastøy, there is a tendency for the media to be vague and reference bullshit statistics without citation. As best I can tell, about 14% of men and 2% of women who serve their time in an APAC facility end up being re-convicted after 5 years. The baseline for Brazil as a whole is closer to 40%, though estimates vary. Critics may argue that this is purely the result of cherry-picking the best-behaved criminals, but the escape statistics do not paint a picture of perfect selection. It’s clearly more than Bastøy can muster, either in terms of relative improvement or absolute recidivism.

Utopian Prisons

As explored in my second essay on Adjudication, I believe Utopia rejects the death penalty, torture, and slavery without exception, but sees “imprisonment” as a suitable punishment for criminals on the grounds that it is an extension of a land-owner’s right to prevent trespass. Just as anyone can kick a visitor out of their house, and local governments can exile people from a country, the world government can effectively “exile” criminals from “the entire world.”12

These people, whom I will here call “exiles,” are only allowed to live in spaces that have explicitly opted into accepting them. I’ll call these spaces that accept exiles “graceholds.”

Anyone, including an exile, can run a gracehold, as long as it meets a few minimum standards imposed by the government:

• Walls — All graceholds must have a strong border that prevents exiles from leaving without an escort. An open futures market for the gracehold must anticipate less than a 1/36 annual chance of any escapes, and graceholds are held legally liable whenever escapes occur.

• Library — All graceholds must have a library holding the Utopian canonical texts on-site, staffed by a librarian (who may themselves be an exile). Exiles living in the gracehold always have the same library access rights as anyone, including the right to private reading time.

• Police — All graceholds with at least two people living there must have some form of police presence to enforce the universal law. This typically involves some of the exiles in the gracehold being employed as sentries — trained to watch for and report crimes. Gracehold police have the same anti-abuse mechanisms (e.g. body cameras and third-party reviews) that the normal police have.

• Mail — All graceholds must permit exiles to send and receive private letters (not packages) as part of the standard global mail system, both on physical paper and electronic mail.

Graceholds do not need to have a policy of universal acceptance (though some do). If a particular exile is violating the rules of a gracehold, the owner is free to cast them out. Some graceholds are more like private clubs or estates, set up by wealthy criminals so they can live out their time in exile in luxury. There are Utopians who complain about the injustice of this, but the consensus is that it’s better to maintain the ethical high-ground than it is to give in to the desire for retribution.

Utopia takes the principle of normalization to the extreme. While some resemble our world’s prisons, most graceholds are closer to relatively normal communities. Most exiles live in private rooms or houses, own property, buy things at normal market prices (both locally and online), do relatively normal work, and live alongside members of the opposite sex (both other exiles and non-exiles).

Smaller graceholds tend to be good for criminals who need to build community relationships, sometimes forming bonds of friendship that last a lifetime. Larger graceholds can be more like microstates, holding thousands of people, most of whom aren’t exiles. These towns allow convicts who have demonstrated reform to live relatively normal lives.13

All of this can happen in a healthy way thanks to consent. Graceholds have the ability to pick and choose which exiles they accept, including by being able to evict them at a moment’s notice. On top of this, graceholds are incentivized to share information with each other about convicts, meaning that a single black mark at one gracehold can mean a world with vastly fewer options. This incentive landscape helps foster good behavior among the exiles.

And on the flip-side, it is technically a misnomer to say that exiles are prisoners because they always retain their universal right to emigrate. If a gracehold is being abusive (or even just being uncool), an exile can decide to move elsewhere. Moving is expensive, both in money14 and in the need to form new relationships, but because all exiles retain their universal right to a basic income, and graceholds can charge rent, there is usually a market for well-behaved exiles, somewhere in the world, at least.

While exiles that have economically valuable skills can sometimes afford to live near their homeland, it is typical for exiles to be forced to travel to a distant place as part of their punishment. The core function of a gracehold is not dependent on any location, so they’re often located on remote islands or in particularly inhospitable areas, to avoid having to pay the cost of building and securing walls.

Because graceholds are funded by basic income and the wealth/income of the residents, Utopian “prisoners” don't impose any additional cost on taxpayers. In fact, because exiles are frequently able to do productive work, graceholds are an important part of the world economy, adding to the wealth of all. These factors mean Utopian courts also experience less pressure for short sentences, and will frequently exile repeat-offenders for multiple decades, even for minor crimes.

In the extreme case where someone is so repeatedly awful that they become rejected by all the normal graceholds, the world government reserves land on a couple dozen tiny, uninhabited, rocky islands to serve as a last resort. Living alone in meager cabins, these worst-of-the-worst criminals are only really able to interact with the occasional mail-boat (often with police defenders) that comes to ensure they have enough food and books.

But people rarely end up in a gracehold of last resort. More commonly, if someone is consistently violent and unpredictable, they will be low-functioning enough that Utopia sees them as someone with mental illness. Justice is blind, and those with mental conditions can still be exiled, if they commit crimes, but they typically wind up in special hospital-like graceholds where their guardians can effectively manage their health and well being.

Prerequisites: None
Foundational to: Part 2

To round out my sequence on criminal justice, I want to talk about the failures and opportunities of prisons. But before getting into my more Utopian views, I want to spend some time appreciating the diversity of prisons that exist in our world. And in particular, I want to consider that some prisons cost a lot more than others.

Bastøy

75 km south of Oslo is the small island of Bastøy, home to a low-security prison housing around 115 Norwegian men.

Bastøy has around 80 buildings, including a lighthouse, a church, and a farm. Notably, it does not have fences or cells. Prisoners are (more or less) free to walk about the island as they wish, interacting with the approximately 70 government personnel who reach the island by ferry, about half of whom are guards. (Only a handful of guards spend their nights on the island.)

In many ways Bastøy resembles a cozy, self-sustaining village — the prisoners chop wood, catch fish, grow food, and care for the shared cottages in which they live, each with their own bedroom. The prison even has tennis courts, a sauna, a soccer field, trails for riding horses and bicycles, and beaches of course. Many inmates spend their days studying in classrooms or training in workshops, building skills that will be needed upon their release. In 2012, the equivalent of the warden said, “If we have created a holiday camp for criminals here, so what?”

The island’s prisoners help operate the ferry which travels between the island and the nearby town of Horten. Trusted prisoners are also allowed to leave the island during the daytime to take classes. On the island, thanks to their various forms of work, prisoners have regular access to knives, axes, and even chainsaws. But despite these risks, almost nobody attempts to escape,1 and incidents of violence are nearly non-existent. Guards are specially trained to treat prisoners with dignity, and inmates know that if they attack someone or attempt to escape, it will very likely end with being transferred to a more traditional facility with tighter security. Prisoners typically need to apply to stay on the island, and while the prison has accepted men who have committed crimes as serious as rape and murder, it selects for those who have proven that they can behave well and are committed to re-entering society.

Bastøy is widely praised across the net for having the lowest recidivism rate in the world2 — 16%! Compare that with a European and American average of 70%! (That second source even says that prior to the 1990s, Norway had a comparable average, and now it’s way down, thanks to sweeping reforms similar to those on Bastøy!)

But wait. What is a “recidivism rate”? The 16% figure for Bastøy is, as far as I can tell, the official Norwegian recidivism rate, as measured by the fraction of prisoners who are convicted of a crime within two years of being released.

It turns out that the widely cited 70% number for Europe/America (and probably Norway in the 80s?) is probably due to shoddy journalism in this 2013 piece from The Guardian! Official US statistics indicate that 68% of prisoners are arrested within three years of being released, but just to spell it out, arrests are not convictions and 2 ≠ 3! In fact, in an apples-to-apples comparison of two-year re-conviction, the rest of the world comes out much closer. Poland is at 28%. America is at 32%. Finland is at 33%. Norway as a whole is 18%.

In fact, when viewed skeptically, isn’t a 16% two-year re-conviction rate… also pretty bad? Like, it implies that even after taking advantage of selection effects and investing heavily in rehabilitation, something like one in six people who walk out of Bastøy basically just immediately commit more crimes upon being released.3 I’m at least surprised that they’re not beating the national average by more than 2%.

Regardless of whether we might hope for an even lower figure, perhaps the decrease compared to the rest of the world is still worth it. The statistics show that Norwegian men are more likely to get their act together, and less likely to continue to harm others, upon release. To evaluate whether that’s worth the cost, we should first ask ourselves how bad it is for a prisoner to be released, only to wind up back behind bars shortly after.

The Cost of Crime

Looking back at my coverage of police and courts, I feel a bit embarrassed that I never actually got into the nitty-gritty of how much a crime actually costs society. This is partly an oversight, but also in part because it’s a difficult question to answer. “Crime” is an unnatural category with uncertain boundaries. How should we count so-called “victimless crimes” that may nonetheless have knock-on-effects that aren’t about the crime itself? And if we’re including knock-on-effects, where does it stop?

Still, we can get a rough handle on things by starting with the direct and quantifiable costs associated with crimes that are some combination of obviously bad and common. Consider the FBI’s Uniform Crime Reporting data from 2019, which aggregates numbers from police reports filed across the country:

It’s pretty obvious from the “value” numbers that they’re considering something like the rough market price of the goods that were taken, and do not include physical or intangible harm. To help broaden our scope, we can consider this 2009 paper by McCollister et al. (hereafter “McCollister”), which has an excellent compilation of cost estimates for various crimes by various authors:

The cost of simple larceny ranges from as low as $344 to McCollister’s high estimate of $3,532. Similarly, motor vehicle theft ranges from $1,723 to McCollister’s estimate of $10,772. Why do the authors consistently have higher numbers? Because they’re factoring in more costs.

Since we’re interested in weighing the cost of crime against the cost of incarceration, let’s focus on harm to the victim. McCollister’s victim+material cost estimate turns out to be broadly in line with the FBI data and the other academic estimates, when considering simple larceny (approx. $700 in 2008 dollars), burglary (~$2000), and motor vehicle theft (~$6300). Robbery is rare compared to the other property crimes, and almost all its damage comes from the violence (and risk of violence), rather than lost property.4

In my essay on adjudication, I cited Scott Alexander’s estimate that each (marginal) year of (American) prison prevents 6.7 property crimes and 1.2 violent crimes. If we assume that all property crimes fall into one of the three categories we have estimates for, and use the FBI data to estimate that 73% are simple larcenies, 16% are burglaries, and 11% are motor vehicle thefts, then the average cost of a property crime is $1524, in 2008 dollars, which is about $2240 when adjusting to 2025 dollars. In other words, if we look at just the concrete harm done to victims, 6.7 property crimes only adds up to about $15,000 in expected damage. We can imagine it’s higher (maybe 2-10 times as high?) when we account for intangible harm, such as distress, chaos, loss of “priceless” goods, and knock-on effects.5

Let’s now turn to considering violent crime. The most well-studied violent crime is homicide, which demands that we put a dollar value on a human life. The authors of the paper that I’ve been drawing on give a tangible cost for murder of $738k, but then bump that up to $9.18M when taking into account the intangible cost of pain and suffering. In 2025 dollars this is about 14 million dollars, which happens to be the value that the US Department of Transportation puts on a life. Let’s go with that.

But of course, most violent crime does not result in death. To get the back-of-the-envelope estimate, let’s also consider robbery, simple assault, aggravated assault, and rape.6 When looking at robbery and simple/aggravated assault (but not rape!), it turns out that the risk of homicide dominates the quantification-of-harm calculus. McCollister, for example, estimates $3300 in tangible cost to the victim of a robbery, which includes medical bills, $5000 in “pain-and-suffering,” and a whopping $17,600 from risk of death (all in 2008 dollars). In other words, 68% of the harm from a robbery comes from the chance that things might’ve escalated and resulted in the victim dying. For aggravated assault, McCollister’s analysis gives an even more pronounced picture, with $8700 in tangible cost, $13,400 in pain-and-suffering, and $81,600 from risk of death — 79%!

(This, incidentally, suggests that policies that reduce the lethality of crime, such as gun control and faster emergency response, might be more effective than trying to directly reduce the incidence of assaults.)

Our last major violent crime is rape, which McCollister estimates as doing $205k (in 2008 dollars) — about $5k in direct harm, $198k in suffering, and $1.5k in risk of death. Interestingly, unlike the property crimes, McCollister is middle-of-the-pack for estimated damages here, probably because the psychological damage of rape is almost never neglected by criminologists. If we take the geometric average of the provided numbers (and a more recent paper) we get an estimate of $157k,7 which is good enough for me.

In a similar way as before,8 we can estimate that 1.44% of violent crimes are homicides, 11% are rapes, 25% are robberies, and 63% are aggravated assaults. If we adjust up to 2025 dollars, the expected cost of a violent crime is around $332k, meaning the 1.2 violent crimes prevented by a marginal year in prison saves $400k, even if we ignore subtle, secondary effects. This is big enough that the property crimes are a pretty minor slice of the pie (~3.6%) by comparison.

Violence is really bad!

Prison Prices

If a marginal year of incarceration prevents around $400k in violent crime, plus maybe $15-30k in property crime, then basically all prisons are clearly worth the expense—for those who might commit violent crimes and be out in public otherwise, at least. For low-level offenders who are responsible for nonviolent crimes like theft, vagrancy, or drug dealing, the net value is less clear.

Norway spends around $160k per prisoner per year (pppy) — close to the highest in the world. Those teachers and workshop tools don’t pay for themselves. For comparison, California spends an average of $128k pppy.

Interestingly, Bastøy is significantly cheaper than the Norwegian average.9 Part of this is because it is a minimum-security facility, and those don’t require as many guards. The typical Norwegian minimum security prison only costs about $93k pppy. On top of that, Bastøy has the advantage of being an island, meaning there isn’t as much security infrastructure to maintain, and of being a farm where the prisoners contribute labor in a variety of ways.

But it’s possible to get much cheaper. Poland spends closer to $12k pppy.10 Texas spends $27k pppy. At the extreme are places like India, which spends only around $416 pppy — about a dollar a day. (And that’s on average. There are many places in India where the average spend is less than half that.)

It’s important not to read too much into these differences in cost. California prisons, for example, are expensive in large part due to high cost of labor and land and a bunch of bureaucratic and legal bloat. The most expensive prisons in the world are high security facilities in places like San Marino, Massachusetts, or Guantanamo Bay, where the prison population is too low to provide good economies of scale. And conversely, the cheapest prisons in the world are ones where many prisoners are crowded into the same space, with absolutely no privacy, regardless of whether the prison would otherwise be high quality. (Norway, by contrast, requires that each prisoner have their own cell.)

Is having some modicum of privacy and distance from hardened criminals helpful in Norway’s lower recidivism rate? Probably. The higher ratio of staff per inmate is probably helpful, too. Let’s wave our hands and say that a third11 of the ~$150k pppy gap between Norway and Poland (so, $50k/year) contributes meaningfully to rehabilitation. Is this price tag worth it to reduce the probability of (2-year conviction) recidivism by 10%?12

Suppose that there is a 15% chance that any given crime is a violent crime, and an 85% chance that it's a property crime.13 This means that each crime costs $52k in expectation. Let’s assume that the average re-offender commits two additional crimes before the one that gets them convicted again. This means that over the first two years after release, a 10 percentage point lower rate of criminality would then be ~$15,600 in expected savings.14 Not all criminals re-offend in the first two years, of course, but also the effects on recidivism are going to be concentrated there. Even if we imagine tripling the preventative effect, it’s still not paying the premium of a single year of Norwegian prison rehabilitation.15

It is hard to get a solid analysis16 that says Norway is justified in spending so much on prisoners, at least if we neglect the cost to the prisoners themselves.

Altruism vs Investment

Despite all this, I think it’s still worth it to try to provide a safe, healthy environment for prisoners. Prisons are some of the worst places on earth, and surely even the most hardened criminals deserve some dignity and care. Finding opportunities to do this in a cost-effective way will be my focus in Part 2.

But before I end this essay, I want to directly challenge the idea that, because prisoners are people, and people are worth helping, it’s good that places like Norway spend money to improve the lives of prisoners.

• Many people believe that the suffering of prisoners is a big part of what makes the punishment just.17 My personal feelings don’t agree with this, but I don’t think my feelings are somehow more valid than other people’s feelings. A pluralistic society should put some weight on those who want harsh punishments.18

• But more importantly, there are much better opportunities to do good with the same money. A donation of just $7 can prevent a child from getting malaria. Saving a child’s life and bringing health to hundreds of other children at the same time costs less than $4k. If we consider money spent on prisoners a kind of altruism, it is deeply ineffective.

There are good programs that help prisoners/criminals, especially first-time offenders and youths. But the thing that makes these programs good is that they clearly pay for themselves by reducing prison costs and making people more productive. We should encourage job training for convicts, but we should also make the people getting the benefits of that training (including employers19) pay for most of it.

Prerequisites: None

Before we begin, I’d like to take a moment to borrow an idea from CGP Grey: let’s use the term “autos” to talk about self-driving cars, robotaxis, and other autonomous vehicles. It’s short and descriptive, and isn’t a huge jump from how the word is already used. For the sake of both brevity and to be the change I want to see in the world, I’ll be using it as the main word in this essay. (The word “robotaxi” also seems like a good word, but I’d like to reserve it for specifically referring to an auto that drives people around as part of a business.)

Fully autonomous1 vehicles have been involved in approximately 43 reported collisions per month in the USA since the government started requiring all incidents to be reported in June of 2021. In these collisions there were two with human2 fatalities:

• January 2025: An elderly man, connected to a previous set of hit-and-runs, was speeding through the middle of San Francisco when he crashed and caused a 7-car pileup, killing a young man, a dog, and severely injuring several others. An unoccupied Waymo, stopped in traffic, was one of the vehicles that was hit. Waymo reports that the vehicle’s sensor logs say the man was driving at close to 100 mph (160kph).

• September 2025: At 1:21am near Arizona State University, a Waymo (with its turn signal on) slowed down to yield to pedestrians crossing the road that it was planning to turn onto. An 18-year-old on a motorcycle behind the auto failed to slow down fast enough and rear-ended the auto, spinning off into the adjacent lane, where he was then hit by a human driver. The 19-year-old driver who hit the motorcyclist fled the scene, but Waymo was able to provide her license plates to the police and she soon turned herself in. The motorcyclist died in the hospital. Waymo was found to be not at fault.

Looking at just this data, one might be inclined to say that self-driving cars have never been responsible for killing anyone. But is this the whole picture? What about outside the USA? What about before 2021? When I search for “autonomous vehicle fatalities” the top hit is Craft Law Firm’s report, which Google summarizes as “Of the 3,979 autonomous vehicle incidents that occurred from 2019 to June 2024, there were 496 injuries and fatalities.” Going to the site gives the more-specific quote: “There have been 83 fatalities related to autonomous vehicle accidents…” Craft Law Firm is very clearly analyzing the same government database as I am. So what’s going on?

First, while there have been some close-calls outside the USA, there have been zero fatal crashes involving autos in the broader world, mostly due to America being the front-runner in autonomous vehicle development.

As for the 83 vs 0 discrepancy, we can get a better handle by considering the two fatal crashes that are listed on the site as examples:

• March 2018: A woman was pushing a bicycle across an unlit 4-lane road (45mph speed limit) in Tempe Arizona. Uber was testing out an autonomous vehicle prototype that had a human driver behind the wheel who was supposed to be watching for dangers and taking control in case of emergency, but at the moment in question the driver was watching a TV show on her phone. The Uber struck the pedestrian, who later died in the hospital. Uber was ruled to be not criminally responsible for the crash, but the human driver was charged and plead guilty.

• March 2018:3 A man driving a Tesla Model X on Highway 101 engaged the car’s traffic-aware cruise control and lane following “Autopilot” feature (distinct from “Full Self Driving”), which insists that the driver must watch for hazards and take over sometimes, and set the desired speed to ~70 mph. The driver was playing a game on his phone when the highway diverged in a slightly unusual way, confusing the software and causing the vehicle to accelerate into a concrete barrier, killing the driver. While Tesla fought the case in court for years, they eventually settled with the driver’s family.

These two cases (which, again, I didn’t pick), are both examples of human negligence in supervising vehicles that were not intended to be used in a fully autonomous way. Regardless of whether their hardware was good enough (we’ll get into that later), neither vehicle had software that was sufficiently robust and capable of handling edge-case situations. The companies knew this. It’s why Tesla insists that drivers keep their eyes on the road and ready to take over at all times when using Autopilot, and why Uber had a back-up safety driver in their prototype vehicle.

More broadly, there is a distinction between “advanced driver assistance systems” (ADAS), which are meant to have a human behind the wheel who is paying attention at all times, and systems like Waymo, which are full autos and theoretically capable of operating safely without any human driver. ADAS are extremely common these days, and because they rely on the driver paying attention, can easily be “the cause” of an accident. It is from these half-measure systems, like Teslas (and the Uber prototype), that we get the 83 fatalities. It is genuinely the case that a fully autonomous driving system — one that has been engineered to a standard where it does not depend on human oversight — has never been responsible for someone’s death.

Safety Stats

But what about non-fatal crashes? Just because autos haven’t really killed anyone yet doesn’t mean they’re safe. After all, full autos are new and rare, restricted to a few geofenced urban areas with nice weather and speed limits that make collisions less deadly. To really judge safety, we need to also look at how common autos are, what kinds of conditions they’re being used in, and what the statistics beyond fatalities can tell us.

I crunched the numbers, and by my analysis,4 there are 2,372 unique auto collision reports in the database, of which at least 88% are from these top three companies:

1. Waymo (Alphabet/Google) — 1670 reports — over 167 million miles on the road5

2. Cruise (bought by GM; now defunct) — 304 reports — ~5 million miles on the road

3. Zoox (Amazon) — 115 reports — somewhere around 5 million miles on the road?6

By this estimate, Waymo has around 10 reported collisions per million miles, Cruise had 60, and Zoox has around 20. While my estimate for the Zoox mileage might be off, such that it’s comparable to Waymo, Cruise was clearly worse. In addition to seeing this in the raw statistics, it’s probably worth mentioning an incident in October 2023 when a woman crossing at a crosswalk in San Francisco was struck by a hit-and-run (human) driver and pushed into the path of a Cruise robotaxi. The Cruise auto not only ran her over, but proceeded to stupidly drag her over 20 feet before it pulled over. The woman, thankfully, survived, but Cruise tried to cover up the fact that their car dragged the woman, only later admitting the error and having their permits revoked in California along with paying $612,500 in fines.7

Even focusing purely on Waymo, these safety stats don’t look so great when we compare them to human drivers. Each year in the USA, there are approximately 6 million car crashes reported to police, which result in about 40 thousand deaths. Since Americans drive about 3 trillion miles a year, that’d only be about 2 fatalities in ~150 million miles, which is a low enough number that the “zero deaths” from autos might just be an artifact of good luck, low speeds due to urban settings, and bias in how I’m judging things (e.g. not including the Uber prototype or the ASU motorcyclist). And more significantly, if we do the naive math on collisions, we might conclude that Waymos get in >5x crashes, since humans only produce about 2 reported collisions per million miles.

But, can this really be correct? Waymo released a safety report in October which tells a very different story. It claims a “90% reduction in crashes that cause a serious injury or worse,” “81% fewer injury-causing collisions overall,” and “82% fewer airbag deployment8 in any vehicle crashes.” That's a missing factor that makes Waymos 25 times safer than the naive data suggests! What are we to make of this gap?

First, let’s try to get closer to an apples-to-apples comparison. If I filter down to collisions where “Highest Injury Severity Alleged” is either “Serious” or “Fatality,” I get six Cruise incidents and six Waymo incidents, including the two fatalities I mentioned at the start of this essay. Six incidents in 127 million9 miles is ~0.05 per million miles, which is more than double the 0.02 that the report cites. It turns out that Waymo is throwing out three incidents that are showing up in my data:

• February 2025: At 3:39am, a Waymo in Chandler, Arizona crossed an intersection. An SUV in a dedicated turn lane beside the Waymo continued straight, went up into the sidewalk, crashed into a street light, and spun out into the lane the Waymo was in, leading the Waymo to rear-end them. The driver of the SUV was transported to a local hospital for treatment. The Waymo passengers (despite not being buckled up!10) were mostly fine.

• October 2024: A Waymo in San Francisco came to a stop in a queue of traffic for a red traffic light in the rightmost lane an intersection. A passenger car traveling crossed the double yellow line and crashed into an SUV that was in the left lane, beside the Waymo, pushing the SUV into the Waymo. Someone involved went to the hospital, but the report doesn’t say who.

• May 2024: At 12:58am, a Waymo in Los Angeles was being tested on the freeway. A speeding car tried to pass the Waymo, but didn’t fully leave the lane when doing so, clipped the Waymo’s rear left corner, and spun out into the median. Again, the report doesn’t say who went to the hospital, but I have a guess.

To my eyes, none of these are Waymo’s fault, with the potential exception of rear-ending the car that crashed into the street light. Still, the methodology on Waymo’s report says “this analysis included all collisions, regardless of the party at fault and Waymo’s responsibility,” and Waymo does include the two fatal incidents I mentioned at the start of the essay (as well as a November 2023 incident where a driver ran a red light). I have no explanation as to why Waymo only seems to be counting half the incidents. The report claims that human drivers have a serious-injury collision rate of 0.23 per million miles, so they’d still be able to claim a 79% reduction even with them included.

Let's take a moment to fact-check the 0.23 serious injuries per million (human) miles driven statistic. Waymo says it comes from two papers written by their safety team: Scanlon et al. (2024) did a base analysis, and Kusano et al. (2025) extended it. To get this figure, Scanlon et al. went through government databases for crash reports and restricted their analysis to surface streets (and excluded large trucks). They found approximately 130 thousand serious crashes per year and estimate that 2.14 trillion miles were driven on surface streets per year. These numbers look about right, and account for the fact that urban environments have more crashes, but fewer serious crashes. Naively, this gives a serious incident rate of only 0.06 per million miles.

To adjust this number up, the Waymo analysts do two adjustments:

1. Each collision involves an average of 1.75 vehicles.
   Imagine if you’re counting “handshakes with (at least one person named) Bob” per person named Bob (analogous to auto collisions per autonomous mile) and comparing that number to the number of handshakes per person (analogous to collisions per vehicle mile). Because most people aren’t named Bob, the first number will mostly be the average number of handshakes people named Bob make, but because each handshake requires two people, the total number of handshakes per person will be about half of the Bob-specific number.
   (Interestingly, over 81.8% of the autonomous collisions involved a human vehicle, which means that if we naively assume that each of those involves just one other car (which is an undercount, as evidenced by our first example fatality), then autos are only 70% as likely as a human driver to collide with a stationary object or pedestrian per collision. I think this makes sense, and follows from autos having fewer blind-spots and never being drunk.)
   When Scanlon et al. adjust for this factor, they get a rate of 0.11 per million miles.

2. Kusano et al. do a fancy spacial analysis that tracks what parts of the relevant cities are dangerous and how much time drivers spend in those areas and concludes that if drivers drove on the same streets that Waymos are tackling each day, then the actual incidence rate would be closer to 0.24 per million miles.

After skimming both papers and thinking about it, I trust the adjustment from Scanlon and I don’t trust the one in Kusano. Even if the analysis is correct in some sense, there’s a selection effect in this particular analysis getting applied (and not other analyses that adjust the bottom-line in the opposite direction).

But even after removing the factor of 2 from Kusano, and the factor of 2 from Waymo’s data randomly missing half of the incidents, we get a serious collision rate of 0.05 vs 0.11 per million miles, meaning Waymos probably get in less than half as many serious crashes as humans.

Returning to our earlier 25x gap, if we take the analysis of serious incidents as indicative, then we can explain perhaps a factor of 4 by Waymo’s thumb on the analytical scale, and a factor of 2 from the Scanlon vehicles-per-collision adjustment. But this still leaves a missing factor of about 3x.

The answer is simple: over half of collisions involving human-operated vehicles do not get reported to police. The reporting requirements on autos are very strict, and can involve basic dings, including with stationary objects. Just going on my personal experience, I have way more than 3 times as many unreported collisions compared to ones that are serious enough to involve the police or even insurance.

So, my best guess at the bottom-line, apples-to-apples comparison is that Waymo is juicing their safety stats, and that they only get into collisions slightly less often than human drivers. But, I think it’s also clear when looking at the individual case reports that almost all serious incidents are the result of human error from other drivers on the road! This means the bottom-line of Waymo’s safety report is still basically right, regardless of what’s going on with their analysis — the USA could likely save tens of thousands of lives each year by adopting autos at large scale. It’s hard to say exactly how much safer the roads would be, but if the technology continues to mature, I think autos could very plausibly reduce fatalities and serious injuries by over 80%.

The Lidar Wars

Let’s return to the discrepancy between fully automatic vehicles, like Waymo, and the driver-assistance systems, like Tesla. Just looking at them, a big discrepancy stands out: Waymos have a giant, spinning machine on the top (and other add-ons). The gadget is an omnidirectional lidar sensor — a way of mapping the auto’s surroundings by shining infrared laser light and watching for reflections. Teslas, on the other hand, are mostly just equipped with normal cameras. Elon Musk has said “Lidar is a fool’s errand. Anyone relying on lidar is doomed. Doomed! [It’s] expensive [and] unnecessary.”

Is Musk right? How expensive and/or necessary are the various sensors that Waymo (and nearly every other auto company) adds to their vehicles?

Musk’s quote is from April, 2019. Just one month earlier, Waymo announced that they had successfully reduced the unit cost of their omnidirectional lidar sensor by 90% — from $75k to $7.5k. Since then, prices have dropped a little bit, but not a ton. An estimate for the 5th generation Waymos has a total sensor cost of $9300, the 6th generation in production now are supposedly cheaper, but my guess is it’s around the same. By contrast, the cameras on a Tesla are only a few hundred dollars.

Let’s turn to whether lidar is necessary. Unfortunately, it’s not really possible to compare Teslas and Waymos for a bunch of reasons. Most fundamentally, any gap in the safety statistics might be from a software edge, or even just a difference in the basic characteristics of the vehicles, like braking ability. But also, Tesla “full self driving” requires a human driver behind the wheel, and that introduces a cluster of complications, such as a bias where the driver is most likely to engage the system in contexts where they think it’s safe, or where the miles-per-accident is lower than average (such as the freeway). When the user is forced to step in due to a critical disengagement (~360 per million miles), would the car have crashed, otherwise?

Tesla’s own report says 0.2 “major” collisions per million miles. Unfortunately they don’t use the same notion of severity as either Waymo or the government data we were looking at before. And we can’t easily check it against the government data, because that data doesn’t distinguish between the premium FSD software11 vs the primitive “Autopilot” included for all drivers. In theory we could check Tesla’s experiments in deploying robotaxis in Austin, Texas, but that program is just a few months old and doesn’t have any public statistics.

What we can do is compare that to their estimated human rate of 1.4 major collisions per million miles, giving approximately a 7x reduction — right in line with Waymo’s claimed safety benefits. For minor collisions the story is the same. But I notice that the report claims that (human) Tesla drivers are between 1.4 and 3.2 times safer than average. This is… false. Tesla drivers have the highest accident rate of any brand of car in the USA.12 No explanation is given for the discrepancy, and in general, Tesla has a poor record of accurately reflecting reality.

Let’s zoom out. The average vehicle travels 156 thousand miles in its lifespan. If we estimate ~$500B in total damages from collisions per year in the USA, then combined with the 3 trillion miles traveled per year, we can calculate that the typical vehicle deals $26,000 in expected damage. Suppose that vision alone can reduce the accident rate by 60%, and with lidar it can go down 85%. That’s an expected savings of $6,500 from having a lidar sensor.

In other words, despite Tesla’s sketchy statistics, I conditionally agree with Elon Musk. If you can reach human parity purely through software, which I see no reason you couldn’t, this could plausibly near-eliminate drunk, distracted, reckless, and sleepy driving — the significant majority of damage sources. While lidar has some promise of making vehicles even more safe, we either need to learn to make it significantly more cheaply (perhaps by switching to solid-state devices) or we need to value human lives/safety more highly for it to be worth it in the long run.13

In the short-run, however, I think Tesla is almost certainly going down the wrong track. Regardless of where the technology is at, and whether they have an AI that can drive from camera data as well as a human, there is a huge amount of social resistance and fear around self-driving vehicles. Autos have literally killed zero people, but still people are working to restrict their deployment.

The first time a Waymo is genuinely at-fault in a car crash, the costs will be far greater than just the loss of life and damage to property. Waymo’s reputation will suffer a huge blow, and with it will come a massive obstacle to the continued rollout of a technology that is actively saving lives. Given this environment, it is wise for companies to err heavily on the side of safety and caution, even if it’s a little pricey.

Festina lente.

Utopian Autos

Urbanists love to criticize US car culture. The solution, they say, is public transit — trains, busses, ferries, and trams — plus paths and sidewalks, of course. I agree that these are great ways to structure dense cities. But even in places with greater investments in public transit, like the densest parts of Tokyo or Hong Kong (two of the least car-centric cities), roads and trucks are still a vital part of last-mile logistics — stocking stores, picking up trash, and facilitating construction work. And as long as these roads exist for buses and essential services, it also makes sense to use them for private car rides.

Outside the dense urban core, cars become even more essential. But the average vehicle stands idle over 95% of the time. These unused vehicles take up space, clogging roads with street parking and consuming 22% of major US cities with sprawling parking lots and expensive garages.

In Utopia, things are more efficient.

Thanks to widespread numeracy, and greater levels of technological enthusiasm, investment capital, and overall wealth, Utopia has a road network where human drivers are only a small minority. Even in rural areas where parking is cheap and access to cars is essential, Utopian citizens default to hiring robotaxis as-needed, rather than owning a vehicle themselves. Robotaxi providers form long-term relationships with customers, and provide guarantees that commuters can get picked up without delay at known points in their daily routine.14 Thanks to this default, children, elders, partiers, and other people who have difficulty driving, have a higher level of mobility than in our world. And thanks to having lots of autos available in off-peak hours, people who travel at unusual times of the day often pay very little.

Having far fewer humans operating deadly machinery means the roads are also safer to cyclists and pedestrians. Utopian autos take a little longer to get from place to place than vehicles in our world, in part due to better adherence to speed limits. But, thanks to not having to pay attention to the road, these longer commutes actually involve more life lived.

Prerequisites: Technically none, though I build on many previous ideas, especially Advertising and Surveillance.

One of the most important ideas in all of human history — perhaps the most important idea — is “the beliefs, aesthetics, and worldview of a person should be cultivated through contact with reason and evidence, not on the basis of authority, tradition, superstition, or dogma.” In other words: it is good to be a freethinker.

Freethought is so centrally important because:

1. Its fundamental nature makes it relevant to nearly everyone. Even young children can learn to notice an argument from authority and distinguish it from something that makes sense according to their own perspective.1

2. It is causally upstream of a huge number of healthy practices and principles. Freethought turns the mind in on itself, seeking a solid foundation for ideas, and leaving people hungry for methodology and rigor.

3. It is so rarely the norm. Bias and lazy thinking seep into even the most careful thinkers and societies. Where and when freethought fades out, humility and collaborative truth seeking go with it, leading to dominance contests and tribalism.

The track record of freethought should be obvious; science is a project of freethought, explicitly developed by freethinkers; liberalism and the associated progress in civil rights have the same pedigree. Transformative progress stems, at least at first, from individuals willing to stray from the dogma of their societies and chart new paths. Tyranny is founded on its absence. And when the spirit of progress turns into revolutionary fire and bloodshed, whether in France or Russia or China, freethinkers stand against the tide, holding onto the truth even at great social peril.

Most of my focus in this essay will be on free speech. This is because thinking, in practice, is rarely a purely private activity. We think through speech — in conversation with those we are close with, in writing out our perspectives, and even in the back-and-forth dialectic with those who see things differently.

It is in the arena of speech rights where freethought often faces its sharpest challenge.2 Those who wish to control beliefs begin by restricting what can be said, written, or heard. Speech is the exposed surface of thought, and so it is where power presses hardest.

But just as freethought cannot thrive without free speech, the project of navigating the various edge-cases of free speech is lost without freethought as a guide. It grounds the defense of expression in something deeper than procedural liberalism and provides a way to navigate the genuinely hard cases where an unreflective defense of free speech can, ironically, become dogma that undermines the inquiry it was meant to protect.

Words and Deeds

Certain things should not be said. Indeed, certain ideas should not be entertained. Pernicious falsehoods, like the notion that standard vaccines are dangerous,3 can spread like wildfire and prove nearly impossible to overcome. Even private thoughts can cause harm—coveting your neighbor’s wife is the sort of thing that leads to trouble, even when the trouble is merely an unnecessary “second arrow” of suffering in the mind of the coveter. The impulse to control speech and thought is a response to real problems.

But even granting that some speech causes harm, suppression is remarkably difficult, especially when speech is most potent. We can see this in the problem of organized crime: imprisonment can prevent a boss from personally committing crimes, but his orders still travel through the prison walls. Indeed, if speech were truly controllable, tyranny might have swallowed the world long ago.4 When certain topics are made taboo, the mind learns to speak in codes and innuendos.

Thoughts are even harder to stop. Even when someone, at a conscious level, wants very much to avoid thinking about something, it can be a struggle to avoid fixating on it. Indeed, the deliberate suppression of thoughts — and words — can perversely make them more likely to occur.

The key property that makes thoughts and words so difficult to suppress is that it’s impossible to determine whether they are permissible in advance of them being generated. We obviously don’t want to suppress all speech, but the boundary between acceptable and unacceptable speech is impossibly complex and constantly shifting. An idea which is harmless on Monday may become toxic on Tuesday, so there is almost nothing that can be done except impose arbitrary and inconsistent punishments on speakers retroactively.

Actions are usually different.5 It is not hard, in advance, to know whether breaking into someone’s house to steal their stuff is an act that society should allow.6 And, when armed with the capacity to think things through and discuss them in a free manner, we can be increasingly confident that rule-breakers have knowingly gone against society. By cleaving to clear principles and setting bright lines, we become civilized.

Reality is not made of discrete categories. Some actions are more like thoughts, and some speech is more like taking an action.

Are there any clear principles that we can use to set a bright line around certain forms of speech that are particularly bad? Yes! And it is precisely the action-like forms of speech — as opposed to the thought-like forms of speech — where it is possible. Spamming, non-consensual advertising, and harassment do not facilitate freethought; they are a form of attack. Doxxing and otherwise violating people’s privacy likewise make it harder for everyone to speak freely and discuss things in a relaxed way. In each of these cases there is work that needs to be done to figure out exactly where to draw the line, but it seems clearly better for the sake of freethought (and general quality of life) to draw a line somewhere.7

The Right to Listen

The standard framing of free speech centers on the speaker: I have the right to voice my dissent. But when the Catholic Church maintained its list of banned books, the Index Librorum Prohibitorum (only abolished in 1966!), the primary victims were not the authors, most long dead, but those who were denied the opportunity to read these works and decide for themselves what to believe.

This reframing is crucial. If we focus on the speaker, we end up debating who “deserves” a platform. And when the censor can cherry-pick the worst of the worst, it’s hard to stand up in their defense. But if we focus instead on the listener, the question becomes: are you sure they don’t benefit from being able to hear?

But the listener-centered framing raises the obvious objection: not all listeners are well equipped to engage with everything that people want to say. Children are not just harmed by some pornography8 and graphic violence, but also by some “pure ideas” such as the prospect of an eternal hell. Adults, too, can fall prey to conspiracy theories and religions that produce “trapped priors,” immunizing themselves against counterevidence. Misinformation often spreads faster than corrections, being built on exciting and outrageous lies instead of boring, nuanced truths.

The Church understood this well. The Index was not titled “Books We Forbid You to Write” but “Books We Forbid You to Read.” Censorship targeted listeners, not speakers—and this was no Catholic peculiarity. The consensus across nearly all societies, until quite recently, was that ordinary people could not be trusted to encounter dangerous ideas.

While the free speech absolutist might wish for a total absence of regulations or restrictions, dogmatic adherence to freedom can simply lead to people being imprisoned in a subtler way, adrift on an ocean of confusion and lies. And yet, if we deem ourselves benevolent protectors — whether government censors, social media policy execs, or simple parents — are we not giving up on the very foundation of freethought?

To take the high road, and avoid either kind of failure, we must adopt better methods. Consider Community Notes as a starting point. When someone goes on Twitter/𝕏 and posts something wrong, other members can augment that bit of writing by providing a correction and/or additional context. Thanks to some algorithmic magic behind the scenes, the note that is most unobjectionable and well-sourced will get the spotlight.

Community Notes isn’t perfect, but it exemplifies the principle of augmentation over suppression, sometimes known as the counterspeech doctrine. On a level playing field, “fight bad speech with better speech” can be a naive strategy — it takes far more work to clean up a mess than to make one. But with clever design decisions, like those seen in Wikipedia and Polis, we can make it easy to selectively highlight and amplify accurate content that builds bridges towards consensus. By making these platforms the strong default, but not a strict monopoly, we can cultivate wisdom while also empowering people to listen to whomever they want.9

For the listener, being able to explore the ideas of others without fear or punishment is vital, even if those ideas are wrong or illiberal. Engaging with compelling falsehoods in good faith, when someone is ready, is one of the best ways of learning to notice that irrationality is not a problem of dumb crackpots, but an endemic and pervasive risk to all people. No authority can teach the skill of being skeptical of authority. By its own lights, the love of freedom must be earned organically.

Utopian Freedom of Speech

Utopia has a strong norm of keeping almost all laws that govern speech, either protecting it or censoring it, out of the hands of the world government, and firmly at the local level. There are places in Utopia where people are not allowed to speak out in various ways, such as committing blasphemy. Such pockets of strong authoritarianism are rare, but almost all local governments have restrictions on advertising, publishing photos of someone without their consent, and on who can access what kinds of sexual media. Somewhere in between, in terms of frequency, are places that restrict access to ideas and thinkers that are considered to be harmful for one reason or another. Almost all these restrictions, however, are mediated by licenses — people are allowed to opt-in to things that could hurt them as long as they can prove that they’re competent enough to navigate the risks.

Strengthening the freedom of individuals, the world government does protect the right to emigrate. If a citizen finds their local government too censorious, and wants to have a taste of what has been forbidden, they only need to leave and go to a place where no such prohibitions exist, including by buying their land from the country that formerly held it and thereby seceding (perhaps with a few of their neighbors), or by traveling out into the many unincorporated spaces.

The one major gap and exception to the world government’s hands-off policy is the indoctrination of children. While children theoretically have the power to travel and even secretly request being adopted by new guardians, they rarely exercise these rights, even when oppressed, due to not having enough knowledge and experience to understand the alternatives or the magnitude of the abuse. Because of this, Utopia protects freethought, in the young and the old, by establishing a universal right to library access.

The futarchy is in charge of choosing the civilizational canon: a set of 1296 books, written in the global language. Once a book is chosen, it has a shelf-life of 60466176 rests, or about 587 years, during which it cannot be removed from the canon, even by a government backed by a supermajority of the populace. Only once its time is up does the futarchy select a book to take its place, perhaps re-selecting that same text to have another tenure. Despite the long timeframe, due to the number of books in the canon, a book is updated roughly every 165 days.

The right to library access involves being able to not only lay hands on any of the canon texts within no more than a ride’s distance10 (~55 km) and a visit’s wait-time11 (~3 hours), but also to learn the global language (both by accessing learning materials at a library, and by speaking with the librarians) and to be able to spend at least 1/36th of one’s life alone in the privacy of a library room, with no one watching to see what is being read. Most Utopian libraries have more than the canon texts, of course, and the shape and nature of such libraries varies greatly from place to place. Many Utopian police and fire stations are libraries, as are most churches, schools, and hospitals. For the Utopians, libraries are something of a sacred space — a place to recognize and honor the power of ideas and the freedom to read the words of one’s ancestors.

On the whole, Utopia stands as a place with vastly more freedom of expression. There are no globally enforced copyrights, thanks to prize-based funding of public goods. There aren’t even universal laws about publishing materials that are near-universally regarded as obscene or dangerous, such as instructions for producing engineered plagues or other weapons of mass destruction. Instead, Utopian law enforcement strives to use above-board methods to notice who is putting out such materials and charge them for criminal actions.

Utopian Freethought

More broadly, Utopia sees the laws around freedom of speech less important than the underlying culture. It doesn’t matter whether you’re legally allowed to give your thoughts a voice if the people around you will condemn you for it. Utopian culture thus celebrates the influence of freethinkers and emphasizes that it is honorable and good to speak falsehoods, as long as they are genuinely believed and one shows good-faith and openness when they’re challenged. We all have been seriously wrong at one time or another, and giving a voice to those wrong thoughts is the first step towards changing one’s mind for the better.

This culture of fallibility is an antidote to the brittle tightness of authority. Experts, politicians, and celebrities in Utopia will often make a big deal out of big ways that they’ve been wrong and have changed their minds. Being a devil’s advocate is encouraged, as is admitting ignorance or confusion, when appropriate. Utopians tend to judge intellectual impressiveness on the basis of rational validity and clarity, rather than on which particular positions someone espouses.

Much like our world, internet-based media, both social and professional, is the “town square” in which ideas are usually debated. The serious people of various Utopian institutions (including, but not limited to, local governments) cultivate freethought in these spaces by offering prizes to the platform-builders who prioritize open algorithms with strong defaults that encourage social harmony and truth, often by providing decentralized ways for society to augment bad content with corrections. And thanks to various government standards, the software that people use to explore and engage with the world of ideas can easily be customized to match their needs and avoid monoculture.

Prerequisites: Carbon Taxes

Suppose you just finished a soda and want to dispose of the aluminum can. You spot a trash can, but something inside you thinks that perhaps you should recycle it. Which is more valuable: the convenience of just throwing it away, or the economic and environmental benefits from recycling?

Ideologues may jump in with an opinion, but a major theme of this blog is to actually do the math. Whether the cost of something outweighs the benefit is not a question that can be answered without looking at the real world and thinking about what’s achievable in practice.

Of all materials, aluminum is widely considered to be the most obviously beneficial to recycle. Cans and similar objects can be recycled for just 5% of the energy that it takes to mine an equal amount of metal from the ground. But energy is only one part of the cost of producing sheets of aluminum ready to be used in factories. What is the actual, bottom-line value of the can in your hand?

Because the exact price varies by quality, location, and the whims of the market, the US market price for scrap aluminum can range from as little as $220/tonne1 (in Oregon) to as much as $4409/tonne (in California). For the sake of having a single number, I think a price of $1500/tonne is a high, but still fairly reasonable estimate.2 If a single empty can is 14 grams, then the estimated price is about two cents. Not very significant!

Maybe you should just throw that can in the garbage…3

But wait! The value in recycling a can is not just in the price of the metal. We also need to consider the more general cost of throwing something away. Constructing and managing a landfill isn’t trivial, either in basic economic costs or in terms of potential damage to the environment and impact on future generations.

The Costs of Landfills

To estimate the costs of throwing things in the trash, let’s start with the “tipping fee” that landfills charge — the market price for throwing things away at the dump. In 2024, the average fee in the USA was $69/tonne, with prices as low as $35/tonne in Mississippi and as high as $151/tonne in Missouri.4 If these prices reflect true costs, then throwing away a soda can in the most expensive landfills in the country might cost as much as $0.002 — just two tenths of a cent.

But perhaps the true cost is higher. Many landfills are run by local governments, raising the questions of whether they’re being subsidized by taxpayers. When weighted by capacity, however, it turns out that the majority5 of municipal solid waste (ie trash) is handled by private companies. In general, privately owned landfills do tend to charge more, but only on the order of 134% of what public landfills charge. Even if we apply this surcharge, the overall price of disposing trash remains quite low.

As a quick aside: consider that the average US resident produces 0.8 tonnes of garbage each year, or about $56 in tipping fees. I don’t know about you, but my garbage bill is way higher than that. It turns out that most of the cost of garbage comes from buying trucks and fuel and paying workers to collect it (along with management and other forms of overhead). The costs from storing the garbage in a safe way is usually less than a third of the total.

Let’s move on to environmental concerns. Might our low dumping costs be a reflection of how we’re forcing future generations and innocent bystanders (including plants and animals) to pay the price in the form of polluted groundwater and potent greenhouse gasses?

Indeed, this is a serious issue. In the 20th century, rapid population growth and urbanization meant that huge landfills began to emerge across the world, and in most places those landfills were entirely unregulated. By the 1970s and 80s, the damage that these sites were doing was undeniable, and countries began to pass legislation to require landfills adopt various protective measures or shut down. Some new sites were constructed, some of the existing large sites were able to adapt, but the majority were closed off to be managed and cleaned up by various government organizations.6

This crackdown on laissez-faire dumping was good.

Consider the impact on groundwater. Rain and other water that moves through the waste creates leachate, which can be loaded with heavy metals such as lead and mercury or other toxic chemicals. This sludge can then spread into the surrounding water table, poisoning fish, birds, and other organisms as it moves through the food web. People who drink the water or eat contaminated plants and animals can then suffer serious illness, especially children, including cancer and lead poisoning. Dumps that try to save space by burning trash can be even worse, pushing toxic fumes and particulate out into the surrounding area, causing all sorts of respiratory diseases. The exact costs are hard to quantify, and depend on the specifics, but a very rough estimate7 is that one urban dump (in the absence of environmental protection) can cause upwards of tens to hundreds of millions of dollars of damage per year.8

The good news is that these dumps don’t really get used anymore, at least in wealthy countries.9 Modern environmental codes require landfills to be isolated from the water table by dense earth and/or specialized liners. Pumps pull leachate from the bottom of the landfill into water treatment facilities, where the pollutants can be removed by filtration or simply by letting the water evaporate/boil off. And while some modern landfills burn trash (or more usually, they burn the methane produced by bacteria), these furnaces are also designed to filter out particulates and prevent anything worse than CO₂ from escaping into the air.

Here’s an excellent video that gets into the technologies involved:

While these environmental protection technologies and practices have a cost, that cost is, in most cases, already included in the tipping fees that landfills charge. (Dumping often used to be free.) After a landfill closes, the land in question can usually be repurposed as a park or other useful purpose, and the world is nowhere near running out of land that’s suitable for dumping.10

Remaining Environmental Costs

There are only two serious environmental concerns that remain for modern landfills:11

• Landfills emit greenhouse gasses, most notably methane from bacteria that eat food scraps and paper.

• Many PFAS, sometimes called “forever chemicals,” are difficult to remove through traditional water treatment processes, and can also spread via the air.

Let’s start with methane. While atmospheric methane only lasts for about a decade before breaking down, during that decade it is such a radically potent greenhouse gas that it will trap ~30 times the amount of heat as the same mass of CO₂ would over an entire century. Many landfills collect and burn their methane, thereby converting it to the much less potent CO₂ and generating a nice amount of energy in the process, but even the most advanced landfills still produce a large amount of “fugitive methane” (especially from food scraps, which decompose quickly).

As I discussed in my essay about carbon taxes, greenhouse gas emissions, in theory, can be dealt with using conditional government bonds, where the polluter is forced to pay some amount of money at the time when they emit the gas, which they can later get back iff the global temperature doesn’t go up (thus incentivizing research into solutions for global warming). For methane, I believe that the short lifespan of the gas means a series of bonds on the scale of years (not decades) is appropriate.

Suppose that a Pigouvian taxation scheme was put in place, forcing landfill operators to pay a fair price for their methane (and CO₂), which could then be paid back to the world. How much would this add to the tipping fee?

This paper estimates that a tonne of municipal waste emits on the order of 0.026 tonnes of methane (after accounting for about half of the methane being captured and burned). It’s difficult to quantify exactly how bad greenhouse gas emissions are, but if we price methane at 30x that of CO₂, and price a tonne of CO₂ at $1000 — the upper-end of what current rates of artificial carbon capture cost12 — then we can give a rough upper-bound on a tax of $780/tonne of trash. A more optimistic number might be an order of magnitude lower.

Can we do the same thing for PFAS?

While initially thought to be inert and harmless, many of the (less-polymeric) chemicals under the PFAS umbrella, such as PFOA (previously used in making Teflon), have been demonstrated to be carcinogenic, disrupt the endocrine system, and cause other health problems. Many of these chemicals have been discontinued by US manufacturers, but similar ones have taken their place. And while the long-polymer PFAS chemicals used in durable products like non-stick pans are probably safe to use,13 many temporary products, such as nonstick food wrappers (ie microwave popcorn and coffee cups), cosmetics, and food itself can contain worrying levels of short-polymer PFAS. Ideally you won’t get tons of these chemicals in your body, but by their temporary nature, they easily wind up in landfills, and due to the hardy nature of the “forever chemicals” combined with the fact that most landfills were constructed before the dangers of PFAS were known, these pollutants can then escape into the surrounding water and air.

It’s hard to know just how dangerous PFAS are, especially since they’re a class of chemical rather than a single substance, and some members of the class are reasonably safe. Worse, we know that these chemicals build up over time, and their health consequences might be subtle and slow. Short-polymer PFAS can travel on the wind and spread across the world; PFOA and other PFAS have been found all over Antarctica, for example.

Still, like with greenhouse gasses, we can form a rough estimate for the costs by considering how expensive it is to do large-scale water filtration (including on leachate!) and then destroying a comparable amount of PFAS as are being released by the landfill. Alas, while it’s relatively easy to filter out and destroy PFAS from, say, a chemical processing plant, it is far harder to eliminate it once it has been diluted in municipal wastewater. This estimate gives numbers for water filtration between $6k and $40k per gram.

Thankfully, each tonne of municipal solid waste only has approximately 0.01g of PFAS per tonne. About 16% of that escapes within the first year, while the rest stays trapped in the landfill. If our worst-case price on removing PFAS is correct, then this indicates we should perhaps be taxing trash at an additional $64/tonne.

The Calculus of Recycling

It should be noted that recycling an aluminum can doesn’t actually do much to reduce greenhouse gas emissions or prevent the spread of PFAS. But, if my analysis is right, even if it was more like typical trash and we were aggressively taxing the externalities, it would still only cost a couple cents at most to throw away, and be worth a couple cents to recycle. If someone throws their cans in the trash instead of recycling them, that’s less than $0.05 of value lost per can. Even a heavy soda drinker who consumes 5 cans per day, will generate less than $7.50/month.

And, as I mentioned at the start of this essay, cans are one of the most obviously good things to recycle. Where scrap aluminum might go for $1500/tonne, extremely high quality, sorted and refined waste glass goes for less than $100/tonne. More typically, scrap glass can be a net negative to recycle. Paper and cardboard can be a little better than glass, ranging from $22 to $182 per tonne. But plastic is even worse, costing more to recycle than it does to make from scratch.

I wasn’t able to find as much data as I’d like, but what data I did find on materials recovery facilities (ie recycling centers) in the northeastern USA circa 2019 indicates that they were getting only about half as much value in outputs (~$50/tonne) as it cost to process their inputs (~$90/tonne), meaning they needed to charge tipping fees of ~$40/tonne before they’d break even. (Not counting maintenance and fixed costs!)

It’s important to remember, however, that plain ol’ trash definitely has a negative value. A $40/tonne tipping fee is a little below market rate, at least in many places, and if we were better at pricing externalities, the value would be even more clear. As long as we’re just thinking about the landfill vs the recycling center, it’s ultimately still better to recycle, even with the plastic and glass diluting the higher-value materials.

But the costs of recycling are mostly not in the materials recovery facility. Handling recycling means getting more trucks, managing more routes, distinguishing different kinds of bins, and hiring workers to check that people are compliant with the recycling codes. Whether these make recycling a net-positive or net-negative can change from month to month, depending on specifics.

More significantly, recycling creates time and stress for individuals. People make space in their house for two bins instead of one, or perhaps several bins in places where multi-stream recycling is the norm. They negotiate with each other about whether to throw away the entire pizza box, or just the bottom, and recycle the top. They fret about whether this is the week for recycling, or whether that’s next week. They waste their lives cleaning out plastic yogurt containers. These tiny costs are hard to notice individually, but they add up over time and at the scale of a society.

Utopian Recycling

Utopians, thanks to being more numerate and economically aware, do not generally find recycling to be an important thing to spend energy and attention on.

Some exceptions exist, however, especially in dense cities where disposing trash is more expensive due to transportation costs. Industries such as automobile repair that produce known, homogenous waste (like rubber tires or scrap steel) will often have recycling arrangements. Some offices and malls will have bins specifically for aluminum cans. Many urban restaurants and businesses that have a lot of cardboard waste will separate out organics to be composted or burned. At the scales of individual households, however, the economics rarely work out such that it’s worth the logistics of managing multiple bins.

Utopian landfills are almost exclusively privately run and publicly regulated in the form of taxes on the landfill’s externalities. Waste management companies compete with each other to reduce their environmental footprint and then pass on the remaining tax burden to consumers. The average person pays about the same amount for waste collection, however, because the light environmental taxes (which become basic income) are roughly balanced by lower operating costs for trash collectors, thanks to not having to deal with recycling bins.

A notable exception to Utopia’s recycling ambivalence are batteries. Lithium-ion batteries can catch fire in a landfill, causing chaos and endangering lives. As a result, residents of Utopia (like our world) are told to put batteries and old cell-phones in a plastic bag on top of their garbage can. The waste management company then gives the resident the equivalent of a ~$0.20 credit14 on their bill per battery provided, as a way of reducing their own legal/tax/insurance costs.

Prerequisites: Customizable Software

The very first “killer app” was VisiCalc for the Apple II — the first instance of modern spreadsheet software. While some earlier software for managing tables of numbers existed, VisiCalc offered several important innovations, such as an interactive interface with numbers that would be automatically computed according to formulae that referenced other cells. In the two years following its release, sales of the Apple II exploded, going from 35k units in 1979, the year before, to 78k in 1980 and then on to 180k units sold in 1981, with many customers saying that they were buying the machine specifically to use VisiCalc. The humble spreadsheet turned personal computers from a hobby gadget to a business tool.

Over the following 46 years, many innovations have been added to digital spreadsheets — advanced statistical analyses, better formatting, generated graphs/charts, cloud-based storage, multi-user collaboration — but the basic structure has remained remarkably consistent. This is because the underlying concept of organizing data into tables is ancient. Some of the earliest written records are accounting ledgers and mathematical formulae from Sumer (in what is now Iraq) dated at over five thousand years ago.

But despite the rock-solid core design and a strong personal sense of affection, I claim that spreadsheets are bad! Or, to be more specific, the modern spreadsheet software paradigm sits on a bunch of bad design decisions that I claim are not present in Utopia. We can have clean, principled, beautiful tables in so many more places, if only we open our mind to the possibilities.

Essentia Tabulae

Tables are great. Not only are they the most natural1 way to organize discrete information in two dimensions (i.e. the space of the visual field), but they’re even a fairly natural mathematical object with very simple rules. On the most basic level they involve:

1. numbers that are arranged relative to each other

2. a format for each number that says how to render it

A key insight is that these two aspects are both foundational and extremely potent.

For instance, many would say that the ability to put text into cells is important, but text is secretly just a number that's been formatted to look a certain way.2 Currencies, levels of precision, and style choices are all easy to see as formatting questions, but observe that we can also interpret formulae as having a “rendering format” that makes the expression “=2+2” appear as “4.” The value of the cell hasn't really changed — it's simply being interpreted in a clever way, much like plain text.

Notice, also, that I said the numbers are arranged “relative to each other” rather than “in a grid of rows and columns.” This is because 2D tables are only one specific version of the core idea. Lists, for example, can work like 1D spreadsheets. And in fact, if we roll with the idea that lists are a kind of spreadsheet, we can see tables as a list of columns, each of which is a list of rows (or vice versa3). While it might seem like this violates the notion that cells contain numbers, mathematics tells us that lists of numbers are also numbers, just with a non-standard format.4

This insight also reveals that modern spreadsheets are secretly 3D! The tabs on a spreadsheet doc essentially create a list of tables, where we can think of each table as a number with a particular format. If spreadsheet software acknowledged this point better, it would allow users to go even further, allowing “tabs” to be organized into a table, or a list of tables, or a table of tables, and so on.

Form and Function

The stock answer for “what is a spreadsheet for” is: organization, analysis, and storage of data. But this is blind to several important aspects.

For starters, spreadsheets are used for data entry. This is almost so obvious as to become invisible. Of course we can enter data directly into the table! This has been true since VisiCalc! Imagine how clunky it would be to need to modify a separate file to change a spreadsheet’s data, or use a separate program to change something. Modern spreadsheets even offer tools for controlling how data is entered, such as drop-down menus or validation checks.

But in many instances, data entry doesn't happen in the spreadsheet itself — it happens via form, such as with Google Forms, or a similar piece of software. Most forms are basically just a collection of labeled fields. If spreadsheet software developers were just a little bit more ambitious, especially in allowing styling to make spreadsheets look nicer, I claim that 90% of data entry that currently involves a form could happen directly on the spreadsheet.

This brings us naturally to another neglected use case: data sharing. The main reasons that people don't use spreadsheets for collecting data from others are privacy and security. In all the spreadsheet software that I’m aware of, it’s only possible to restrict visibility or edit power to the entire spreadsheet. This is an obvious oversight. Spreadsheet owners should have the power to mark cells/rows/tables as needing different permissions, either as a whitelist and/or a blacklist. A standard form could then be replicated by granting each unique user (besides the special users like the owner) permission to edit only a single row, and perhaps only for a limited timeframe.

But perhaps the biggest blind spot doesn’t concern neglected features of everyday spreadsheets, but rather the software that was built because standard spreadsheets are lacking. Database systems (queried with languages like SQL) and notebook environments (eg Jupyter) largely arose to deliver speed and scale that spreadsheets struggle to match. Why are spreadsheets so slow? There are two main reasons:

1. Interactivity — Spreadsheet users expect the values of cells to update instantly whenever they make a change, which can mean scanning the entire sheet to update references, recompute summary statistics like averages, and update conditional formatting. Often, a change in one place will trigger a successive series of downstream updates, which can quickly slow down even a modern machine.

2. Flexibility — Spreadsheet cells are typically represented basically as text. Users, for example, don’t need to clarify that “0.17” is a number, rather than a string of four characters, when they type it in. This can get rid of a lot of bureaucracy for the user, but it ultimately prevents the machine from packing cell data into more efficient formats like dense arrays and processing it in parallel. Similarly, functions like VLOOKUP, RAND, and TODAY are powerful, but in conjunction with the interactivity, often result in volatile cells that need to be constantly recomputed.

The great irony is that if spreadsheet developers had decided to allow users to turn off features like instant-recalculation or the ability to insert text anywhere, they likely could have produced a more general-purpose data analysis tool that’s capable of efficiently doing everything from adding two values, to creating 3D visualizations of multivariate polynomial regressions.

Utopian Spreadsheets

My vision for spreadsheets in Utopia starts with a single cell. Each cell has some data, a type, and a set of formatting modifiers (the type and modifiers are collectively called “the format”). Types are drawn from a small set of options, including integers, rational numbers, and strings. Data is always, conceptually, a natural number that defaults to 0. Basic formatting modifiers include making the cell render in a bold font, or with a different color background (modifiers are allowed to be mutually exclusive).

While I’m mostly not going to be focusing too much on interface and presentation, since as part of the culture of customization, Utopia allows a variety of different editing programs with a variety of different interfaces to edit the same basic data structure. But I think it’s worth imagining a portion of the screen describing the selected cell, topped by a drop-down menu for type, then having a field where data can be entered, and then having a collection of buttons, and other toggle options for modifiers that are either on or off.5

The default cell type is a sparse, infinite, bi-directional list. In other words, if we consider a single default cell, we can imagine it to be a list of infinite items, each of which are default cells, which can be seen as lists of infinite default cells, and so on. In this way, when one creates a new spreadsheet, it can be seen as a single cell or as an infinite dimensional array with infinite size in each dimension.

Using modifiers, one can change the default format of sub-cells. In this way, one could re-create a 2D spreadsheet much like the ones of Earth by specifying that the sheet is a list of columns, which are each, by default, a list of strings. But this still won’t be a perfect reproduction — Earth’s spreadsheets are finite size. (Google Sheets, for example, defaults to 1000 rows and 26 columns.) To match this more accurately, one can use another basic type of cell: a dense list. Dense lists have a finite number of elements, and thus a dense list of dense lists will have a “top left corner” which we could call “A1.”

Utopian spreadsheets start indexing at 0, not 1, of course. And nobody would label columns things like “A” or especially “AA” as the column after “Z”.6 Earth’s notion of the cell “A1” would simply be “0,0” in Utopia. But references to cells by using an address like “0,0” are rare, because a common modifier that a cell (and thus a column or sheet) can be given is a name. Thus instead of putting the string “Profit” as the first cell in the column where profits are marked, the entire column is given the name “Profit” and can be treated as having a consistent type (eg currency). To reference the 34th row’s profits, one might write something like “Profit.34” or, if it’s a relative reference within row 34, by something like “myRow.Profit”. Relative references are usually distinguished from absolute references by similar mechanisms.

Utopian spreadsheets default to having very few types and modifiers, but cells can be modified to add more. For instance, there is no type for currency in a default spreadsheet, since currency is culturally specific and might change over time. Instead, to talk about money, the top-level cell of the spreadsheet could be given a modifier that adds “US Dollars” as a type for all the cells contained within. Types are written in a programming language in a way that specifies how they encode and render the data, as well as other features like automatically correcting invalid input (e.g. rounding to the nearest cent). Which programming language is used to specify types? This, too, is customizable. The only constraint is that the type code provide an interface which specifies the rules in a standard way.

Similarly, cells in a Utopian spreadsheet do not have any default way to reference other cells or to be interpreted as formulae. To calculate 2+2, one must first modify the cell to specify the language. For example, one could mark the top-level cell of the spreadsheet as using Excel’s language, which would add functions like TEXTBEFORE, and then mark a specific column to use Google Sheets’ language, giving access to QUERY. More realistically, one can mark specific cells as computational cells that use powerful languages like Python or R, but have the ability to reference the spreadsheet data as though it were an input to the function being defined.

Code that affects an entire spreadsheet can be written as a modifier for the top-level cell. This not only allows custom functions to be defined, but also theoretically gives access to interoperability with other files on either the local machine or over the network. Code can also be sandboxed to particular sections of the spreadsheet and given resource limits, reducing bloat. It is standard when writing a formula to indicate whether it should be re-run manually, or whether it’s fast/safe to re-run whenever the input data changes. These features in concert allow Utopian spreadsheets to handle advanced data analysis and full-scale databases.

Customization naturally adds layers of complexity and fragmenting, but Utopia handles this through standards, rather than limits on the software. If you want to use a fancy new font, go ahead, but also recognize that by doing so you’ll be increasing the file size and potentially making things worse for other users. Businesses and cultures within Utopia gravitate to de-facto loadouts for how spreadsheets get used, but these are in constant competition with each other and go through waves of stylistic refinement as cultures settle on how to use the power of their tools.

We could go over many other features in depth, such as embracing version control to help spreadsheets maintain local undo histories in an efficient way, or the emphasis on functional programming standards and strong typing. We could talk about how Utopian spreadsheets can sort and view data without fundamentally moving it, can hold images as first-class data (including having graphs that are generated by formulae), and of course have basic stuff like the ability to be collaboratively edited in the cloud by multiple users. But ultimately, all these are downstream from the most important difference between how Utopia handles spreadsheets and how Earth does:

Utopia treats spreadsheets as the first-class tool for handling data of all kinds, including for input, heavy-duty analysis, and presentation. Much like our world has a meme of “this meeting could’ve been an email,” Utopia has notions like “this website could’ve been a spreadsheet.” Utopia makes a few tradeoffs in different places, but it mostly achieves superior design by recognizing the primacy of tabular data and pushing the underlying principles to the limit.

Prerequisites: Emotions

Have you ever met a psychopath? I have.1 It was wild.

I met him while traveling cross-country by bus as a young man. He had recently gotten out of prison. (He’s not the only person I met on a bus who had just gotten out of prison, but he’s the only one who was obviously psychopathic.2) He told me — quite proudly — about how he used to work as a banker who specialized in pushing predatory loans on people who couldn’t really afford to buy a house. He said he saw the borrowers as being like prey, and after extracting thousands of dollars from these poor people and driving them to financial ruin, his bank would foreclose on the property and collect insurance money from the government. When the 2008 housing bubble collapsed, he’d been identified as a bad actor (“thrown under the bus” by his colleagues, according to him) and had been sentenced to a couple years in prison. But now that he’d been released, he told me he was excited about the opportunity to get back out into the world and start taking foolish/weak/pathetic people’s money again.3

While this man was unapologetically narcissistic and predatory, he didn’t seem stupid. If anything, he was strangely charismatic, drawing me in with stories of his life. I could very much believe he’d had a high-paying job at a bank and had been able to mask his true nature behind layers of charm when he wanted.

If you look up psychopathy on Wikipedia, it begins with: “Psychopathy … is a personality construct characterized by impaired empathy.” And the entry for empathy begins: “Empathy is generally described as the ability to take on another person's perspective, to understand, feel, and possibly share and respond to their experience.” When read literally, this seems to indicate that the man I talked to couldn’t have been a psychopath, since he seemed very capable of understanding the thoughts and experiences of others — he just didn’t care.4

Instead, I think the right takeaway is that our society doesn’t have a crisp conception of what “empathy” is. Like with emotion, we neglect a core part of what it is to be human, in large part because it’s invisible and internal. Here are several distinct things which “empathy” might mean:

• Social modeling (a.k.a. “cognitive empathy”) — the ability to understand and reason about someone else’s experience.

• Emotional contagion (related to “affective empathy” and “somatic empathy”)5 — the tendency to feel similarly to someone else.

• Sympathy/compassion (a.k.a. “compassionate empathy”)6 — caring about alleviating someone else’s distress.

In neurotypical people, these capacities go together. We consider the experiences of others, and, in the process feel, something like what they feel, and, insofar as that feeling is unpleasant, we seek to help them. But these capacities can also come apart, and not just in psychopaths. It’s common, for example, for people to deliberately avoid paying attention to people who are suffering (such as those who are homeless or living in severe poverty) because it’s a bummer, even if those in need could be helped — demonstrating emotional contagion without (much) sympathy. On the flip side, an Effective Altruist might dispassionately donate their kidney to a stranger — demonstrating abstract compassion without any concrete knowledge of the recipient or having any induced feelings.

Having conceptual clarity is important, not just because it’s good to know what we’re talking about in general, but also because these systems tend to carry a lot of weight in people’s psychology, underpinning everything from morality to theories of self and consciousness. When we lump these capacities together, and don’t examine their limits, we can end up assuming too much, limiting ourselves in various ways.

Empathy’s Origins

The earliest aspect of empathy to evolve was almost certainly emotional contagion. Emotions, as I discussed in my essay on the topic, are an ancient aspect of animal psychology, guiding creatures as simple as worms in the ways of finding food, avoiding predators, and mating. In any sort of setting where animals of the same kind live together, such as a school of fish, there’s an obvious utility to noticing when others around you are agitated or calm, and adjusting your emotional state to be closer to theirs.

As behaviors and emotions became more complex, emotional contagion adapted and expanded. Signs of disgust would invoke disgust. One animal yawning could spread to those nearby. One male acting dominant spurs nearby males to puff up and fight, lest they lose prospective mates. The list goes on.

In mammals and nesting birds we likely have a case of convergent evolution of the roots of sympathy, in the form of taking care of children. It is evolutionarily useful for the distress of a baby to cause distress in the parent, but it could be catastrophic for parents to merely become distressed, since that might provoke all kinds of counter-productive responses, such as running away, hoarding food, or curling into a ball. Instead, evolution selected for parents that would respond to the distress of their children by selflessly attending to the needs of the child.

While all mammals (and almost all birds) take care of their young, it seems likely that these specific impulses to caretaking tend to evolve into more broad kin-selecting sympathy in social animals. Adolescent Florida scrub-jays, for instance, will sometimes stay with their families and help raise their brothers and sisters for years after they’re old enough to go off and have offspring of their own. And beyond the more obvious cases of alloparenting, there are instances of adults helping other adults, such as when vampire bats — who can only go a few days without eating — will share food with others who weren’t lucky enough to find their own.

In parallel, my guess is that social modeling evolved in response to predator-prey dynamics. If a predator can learn to imagine the world from the perspective of its prey, it can track things like whether the prey is alert or calm, and which way an animal is likely to run if attacked. Unlike emotional contagion, which (in many cases) does not need a firm distinction between self and other, tracking the mind of a prey animal requires clearly understanding that its fear does not imply the predator should be afraid. And while there is an advantage to prey that can model predators, the resource investment is asymmetric — a fox preparing to hunt can dedicate its full attention to modeling how a rabbit will run, while a rabbit that spends similar energy trying to model foxes (and eagles and all the other threats coming from all angles) throughout its day will exhaust itself well before any predator appears.7

In the evolution of predatory animals with social structures, like wolves, lions, dolphins, ravens,8 chimps, and humans, the capacity for modeling prey likely expanded and combined with the neural structures used for taking care of kin. This more psychologically complex conception of other members of the same species likely unlocked things like advanced hunting strategies, imitative culture (including tool use), deception of other members of the group, and forming long-term alliances with non-kin. (Alliances, being particularly important, were likely anchored by instincts related to caring for children, such as crying and consolation.)

There is reason to think that this rise in cognitive empathy was particularly relevant to the evolution of human beings. As early hominins learned to hunt in packs, and in particular to scavenge carcasses and crack open bones with stone tools to get access to calorie-dense marrow, they unlocked both the resources to grow much larger brains and had an environment where those with the most ability to coordinate and learn were rewarded. As group size increased (in order to share knowledge and face off against other African savanna carnivores), early hominins who had the ability to deceive and manipulate others and to identify and punish the untrustworthy had an evolutionary edge, kicking off an intelligence arms-race. All of which would be impossible without a rich ability to imagine the world from the other monkey’s perspective, so to speak (along with a rich collection of other factors, such as the huge caloric surplus that came with the harnessing of fire).

Against Empathy?

In his influential 2016 book, Against Empathy, Paul Bloom makes a compelling case that empathy is bad. Or more particularly, he argues that the experience of emotion that is brought on in connection with others, which I've been describing as emotional contagion, is an inferior basis for moral action compared to cold rationality. On the whole, I agree with this thesis, and broadly recommend the book. The spark of mutual-emotion is terribly hit-or-miss, biased, and unprincipled.

But I do think it would be a mistake to broadly condemn emotional contagion. And indeed, even Bloom has some nice things to say about it at the end of Against Empathy. Most notably, he and I share the position that this kind of empathy can be fun, both as a way of connecting with others9 and as a primary way of experiencing art. Harry Potter doesn’t need to be a real person for me to feel “his” excitement about becoming a wizard. Indeed, the word “empathy” was coined in 1909 as a translation of “Einfühlung,” which is the act of connecting one’s emotions with a work of art. And just as experiencing pain can be fun, experiencing sadness, anger, horror, and disgust in an artistic context can be a great way to explore and become familiar with those feelings in a safe context — something that would be impossible without emotional contagion.

But the most important aspect of this kind of empathy, I believe, is guiding people towards caring. While acting compassionately does not require having any particular emotional response, the two are strongly linked. Consider rats that have access to a lever which produces food, but gives an electric shock to another rat nearby. Rats in this experiment express something like sympathy, pushing the lever less often than if it caused no pain to others. This basic response could be the result of abstractly caring about the other rat, or it could be the simple result of emotional contagion causing the acting rat to feel bad whenever it pressed the lever, causing it to behave more compassionately. In a similar way, children experience natural rewards and punishments through emotional contagion that direct them towards helping others. Those who, for whatever reason, don’t have the same emotional contagion, are more likely to be callous and mean. (Notably, those who have normal emotional contagion but reduced social modeling may appear similarly uncaring, but are significantly less likely to deliberately hurt others.)

In this way, emotional contagion is a bit like the training wheels on a bicycle. On a bike, it is difficult to learn to balance without first having speed, and trying to go fast without having skill can be scary. Training wheels help give kids confidence and put them into a context where it’s easier to learn the relevant things. Analogously, emotional contagion allows kids to naturally find their way into win-win relationships where it’s relatively easy to learn to value the feelings of others.

But, also like training wheels, it can be risky to learn to be compassionate and moral entirely through empathy — one can become too dependent on the crutch. Against Empathy does a great job expounding on the risks and biases here. We care too much for those who resemble us, and not enough for those who are alien. Bloom also does weakly acknowledge that affective empathy can be a kind of “moral guide,” but I think a balanced view would give it credit for being the primary developmental tool for compassion, moral thought, and learning to conceive of others as equally conscious and real as ourselves.

Empathy in Utopia

In a more enlightened world, we wouldn't make the mistake of lumping three distinct psychological capacities under the single muddled term “empathy.” The Utopian language has separate words for social modeling, emotional contagion, and compassion — much as we distinguish between sight, hearing, and touch rather than calling them all “sensing.” This linguistic precision reflects a deeper understanding: these capacities serve different functions, develop through different mechanisms, and can be cultivated independently.

Utopia sees social modeling and compassion as straightforward virtues. The ability to understand others’ perspectives enables cooperation, reduces conflict, and allows for the kind of complex coordination that makes civilization possible. Compassion — the genuine desire to help others — forms the bedrock of moral action. Thanks to being more distinct from emotion, Utopia conceives of compassion as less centrally about pleasure and suffering, and has a more holistic and/or pluralistic notion of what it means to help others as ends in themselves.10

Emotional contagion, however, receives more nuanced treatment. Utopian educators teach children to learn how to dial their emotional resonance up or down as circumstances warrant, understanding it as a powerful tool rather than a moral compass, per se. Sometimes it's valuable to feel deeply with others — such as when celebrating a loved one's triumph or experiencing art. Other times, excessive emotional contagion can be counterproductive or even harmful. It’s bad for doctors to feel their patient’s pain. Disaster relief coordinators shouldn’t be paralyzed by the emotional weight of each individual story when forced to triage thousands of victims to save the most lives. Even therapists are better off being understanding, warm, and emotionally unshakable, rather than awash in the feelings of their clients.

Most importantly, children in Utopia learn early that emotional contagion is only one possible pathway towards morality, and one that cannot be fully trusted. They’re taught to think of affective empathy similarly to how we think of folk physics — a useful heuristic that works well enough in everyday situations but which has known flaws. Just as you wouldn’t design a bridge based on intuitive physics, you shouldn’t base moral decisions solely on who makes you feel sad. Feelings are data, not conclusions.

This clarity extends to how Utopia handles its psychopathic citizens. In our world, we conflate the psychopath personality traits with antisocial behavior, creating an unhelpful stigma.11 Utopia, recognizing that compassion and morality can exist without emotional resonance, has developed interventions that help psychopaths integrate into society and find moral wisdom without relying on a one-size-fits-all insistence on emotional responsiveness.

Indeed, some Utopian fiction features psychopathic heroes — characters who are able to make difficult moral choices based on principles instead of tribalism. These stories help neurotypical citizens appreciate the unique strengths that different minds bring to their society, and give psychopaths positive role models to emulate. There is still a strong sigma around predatory behaviors, and an understanding that psychopaths are at greater risk of becoming predators, but with a healthy understanding that nobody is destined to evil, and nobody should be excluded or condemned for factors beyond their control.12

Prerequisites: Part 1, Futarchy

“Due process” has been a major topic of discourse in recent months, mostly because the Trump administration has been heavily disregarding it as part of the crusade against illegal immigrants. The point of having a clear and fair process for deciding who is guilty is that, in the absence of such a process, innocent people — who have done nothing wrong and have never been judged in court — will be imprisoned and suffer.

Unfortunately, absence of due process isn’t the only way in which innocent people who have never been convicted of a crime end up behind bars. At the time of writing, I estimate that the United States has at least 100,000 of these innocents locked up — roughly 5% of the entire incarcerated population.1

The issue here is that when someone is arrested and put in jail, they can be held for months before they’re found not-guilty by the court.2 These people may have done nothing wrong. But due to bad luck, they find themselves suffering a major, unjust consequence at the hands of those meant to bring justice.

On one hand, if the justice system is fast and reckless, innocent people will be imprisoned and the guilty will go free. On the other hand, if the justice system is slow and methodical, innocent people will languish in jail (and the process will be more costly). What is needed is a way to produce accurate and speedy justice.

Plea Bargains?

Do plea bargains help? Trials are expensive and slow. If someone pleads guilty, this frees up the court to focus on getting innocent people cleared and back to their lives. In many cases a criminal knows they’ll be found guilty. Why not make a win-win deal to go slightly easier on them in exchange for helping the system operate efficiently?

In the United States, over 90% of criminal cases are resolved through plea bargains. The system that’s set up right now is not designed to be able to handle more, making bargaining not just a convenient option, but a necessary practice. If we decided to end plea bargaining, as a society, we could, after some adjustment period, presumably hire lawyers, judges, clerks, etc. until our courts had the capacity to handle it.3 But my guess is that this would cost more than three hundred billion dollars per year4 (~$900 per capita; close to half of the total cost of the armed forces). This isn’t an impossible ask, but it would clearly add a large tax burden.

And yet… maybe this cost would be worth it. While a system that is overburdened can speed things along by having people plea out, a system that is well-funded can’t actually be made faster for innocents by having some people plead guilty. And, importantly, plea bargains have a lot of downsides.

• Overly-harsh sentences — The typical incentive for a defendant to take a plea bargain is reduced sentencing, but this sets up a perverse incentive for the government to threaten punishments that don’t fit the crime so that prosecutors can get more leverage and make more deals.

• False confessions — Someone who is unlucky enough to be charged with a crime they didn’t commit might expect to be found guilty at trial, or might expect that even if they’re found innocent, they’ll be ruined by legal fees. Even if such people benefit from taking plea bargains, the deal is done at the cost of justice and their honor, casting further doubt on all others who confess.

• Corruption and bias — Giving prosecutors broad leeway to make back-room deals with potential criminals reduces transparency and produces opportunities for bribery and other corrupt relationships to fester. Even well-intentioned people can end up perverting justice by letting prejudices color their deal-making.

Plea bargaining has its merits, but it ultimately fails to solve the speed-vs-accuracy dilemma, merely papering over it at the cost of justice itself. Though it helps keep costs down, it forces an impossible choice on defendants and grants prosecutors dangerous, opaque powers. We must do better. We need a system that surfaces the truth efficiently and transparently. And to build that system, we must first be clear about what we want our punishments to achieve.

The Aims of Sentencing

Justice serves many purposes. The top five (in no particular order) are probably:

• Retribution — Harming wrongdoers as an end in itself.5

• Deterrence — Preventing crime by giving it a negative expected value.

• Rehabilitation — Helping criminals find a better path, and re-integrating into society in good ways.

• Reparation — Compensating those harmed by unlawful action.

• Incapacitation — Physically preventing criminals from committing more crimes.

These goals can be in tension with each other; retribution and rehabilitation are famously at odds in most situations. Which of them should society, ideally, be focused on? I claim it is incapacitation, primarily implemented through removing criminals from society for long periods of time (a.k.a. imprisonment).

• Retribution is controversial, and naturally cruel. Imprisonment (and the shame of being convicted) should be harsh enough. Torture for the sake of torture is corrosive to the spirit of compassion and mercy that a good society should foster.

• Deterrence is good, but hard to pull off, especially for major crimes. The main reason that most people don’t commit crimes is not because of fear of punishment, but instead because they have a conscience and want to adhere to norms. (Consider the vast number of people who behave lawfully even when it’s obvious they won’t be caught.) Criminals, then, tend to be people who are some combination of desperate and psychologically abnormal, and it turns out that it’s pretty hard to deter such people using hypothetical long-term consequences.6

• Rehabilitation is good, but also controversial and not universally applicable. Some programs show good returns on investment, especially for first-time offenders and youths. Heavily investing in helping those who make a big mistake avoid having their lives fall apart due to prison, and ideally finding a better life path, seems wise. Many people, however, are probably beyond reach, or are otherwise unlikely to be an efficient use of social resources for various reasons.7

• Reparation is good, but not always possible. Often, the damage done cannot really be undone, and monetary compensation only goes so far. Additionally, criminals are usually poor, making it difficult to extract enough wealth from them to serve as genuine compensation. (And such extraction will leave them poorer and naturally more desperate in the future.)

But incapacitation just works. The majority of crimes are committed by a tiny fraction of the population, usually repeat-offenders, and both first-principles and empirical studies both show that removing these people from society is an effective way to fight crime. Scott Alexander’s analysis estimates8 that on the American margin, each year of prison prevents approximately 6.7 property crimes and 1.2 violent crimes, thanks to incapacitation.9

Imprisonment is also a relatively fair way to punish people, compared to options like fees or torture, since people have different levels of wealth and pain sensitivity, but everyone gets roughly the same number of hours in the day.10 Furthermore, if a punishment is framed in terms of time in prison, we can deduct time spent in jail from the duration of punishment. This doesn’t solve the issue of inadvertently locking up innocents, but it might ease the pressure to get a quick trial for those who are clearly guilty.

When Must We Kill Them?

Imprisonment is not the only way to incapacitate people. In some sense, the most effective means of incapacitation is execution. It’s hard to commit crimes if you’re dead.

But at what cost? Let's start with the financial expense. Common estimates for those sentenced to death indicate that it can cost millions of dollars more than life in prison without parole! If it’s cheaper to keep people locked up than it is to kill them, why would we execute anyone?11

Perhaps because it’s excruciatingly painful, and thus serves as a catharsis for those seeking retribution? Opponents of the death penalty often highlight cases of executions that do, indeed, sound excruciating. But a closer look shows that these horror stories often assume suffering without good evidence, or come from botched executions, where one or more things went wrong. The standard method of execution is “lethal injection” — usually an application of a sedative, followed by a paralytic, and then by a high dose of potassium to stop the heart. In theory, the sedative renders the person unconscious, and they never feel a thing when they die.

Very frustratingly, scientists basically haven’t checked this. The only time that I am aware of where brain data was tracked during an American execution was in the execution of Willie Brown Jr. in North Carolina in 2006. During Brown’s execution, he was wired to a Bispectral Index (BIS) EEG monitor. Once he fell below the threshold for unconsciousness he was given the paralytic and heart-stopping injections. Court filings say the monitor never rose above the threshold, but the underlying data was sealed under a court order and has never been released to the public. No state has tried since, probably because of fears around being sued for “cruel and unusual punishment” and/or difficulty in finding a medical professional willing to risk losing their license. With the pro-execution crowd fearing legal trouble and the anti-execution crowd fearing data that clearly shows an absence of pain, all we are left to go on are guesses.

Still, it seems highly likely that, when properly administered, a lethal injection is close to painless. We can infer this by comparison with animal euthanasia, which is far less stigmatized. EEG-based scanning methods indicate that a heavy dose of barbiturates or other sedatives can put large animals to sleep and keep them under, even while their heart stops.

Euthanasia, in short, is a humane method of execution. So it really can't be the case that we kill people because it’s a painful retribution. Some may feel that it’s a fitting punishment for murderers, but I would argue that if retribution is the true goal, years of solitary confinement is far worse than being put to sleep. And it would be cheaper.

But wait. That can't be right, can it? Why is it so expensive to execute someone? Putting animals down certainly doesn’t cost millions of dollars. The Chinese manage to execute people by lethal injection for less than $100!

Indeed, the “high cost” of execution in the USA has almost nothing to do with the intrinsic costs of killing someone, and far more to do with the bureaucracy surrounding death row. Over three fourths of the added cost is from additional legal fees and court costs, including appeals at various levels. Most of the rest is from the additional needs for death-row housing, which holds the average person sentenced to death for over 20 years. If we, as a society, decided to execute people without all the delay and extra effort, it would be far cheaper than incarceration.

But all of that bureaucracy is there for a reason. Consider that since DNA testing became available in capital cases in the early 1990s, at least 34 death-row prisoners in the United States have been exonerated by DNA evidence. That is 17% of the 200 total death-row exonerations on record, and it works out to roughly one DNA-based exoneration for every forty-odd executions carried out in the same period. Overall — counting every kind of evidence — about one death-row exoneration occurs for every eight executions, and statistical modelling puts the underlying error rate at a minimum of 4% of all death sentences. We spend so much more effort because death is permanent, and most people find it barbaric to kill people only to later decide they didn't deserve that fate.

This is a good enough reason, for me, to believe that the death penalty is a bad strategy. But I also think it's worth rejecting on more symbolic grounds.

Civilization is the collective effort to put aside violence as a way of resolving conflict, and to instead turn towards laws and principles that guide us towards virtue. One of the most central and universal laws is the prohibition of murder. Similarly, one of the most important virtues is extending and protecting life. When we decide to collectively kill someone — even a heinous criminal — we are eroding the injunction against murder. We are implicitly saying “killing people is good as long as enough people are for it.”

It's this same principle that leads me to reject forcing prisoners to work, or to perform any act against their will. We have a word for this: slavery. Just as I am against murder, I am against slavery, regardless of whether the government is involved and who is being enslaved.

Imprisonment, by contrast, can maintain the ethical high-ground. If we see prisons not as a place that a criminal must go, but rather as the only place that the criminal is permitted to go, then imprisonment is merely an extension of the right to prevent trespassing. It becomes an “exile” to the only country that will accept them, rather than a violation of their rights to bodily autonomy.12

Utopian Adjudication

In Utopia, when someone is arrested for a (universal) crime, two prediction markets are created and they’re immediately informed of what crime they are being accused. One of these markets asks whether the accused will be found guilty, while the other asks: “Conditional on the judge reaching a verdict, how many days of imprisonment will the sentence be?”

These markets are attached to a public13 court case containing the relevant evidence and other data surrounding the arrest. Journalists (without conflict of interest) are required to be able to shadow prosecutors, making it difficult to suppress evidence or ruin the market with insider trading.14 Indeed, because of incentives that we’ll explore in a moment, law enforcement is heavily pressured to submit all information relevant to the case as soon as possible. (The defense can also submit information, as can relevant third parties such as victims or witnesses.)

The accused is then given a hefty chunk of free liquidity (e.g. ~$5000 of shares) which they can and must spend on the guilt market before open trading begins.15 Their opening position serves as something like a “plea,” albeit without some of the baggage. Someone who bets that they’ll be found guilty, for instance, might simply be saying “I don’t trust the system to realize I’m innocent.” If the accused bets that they’ll be found guilty, however, it’s rare for others to disagree, and the guilt market will usually converge to an extreme “guilty” position.

These prediction markets aren’t just for speculators — they let law enforcement quickly learn whether they have the right person and whether this case is a priority for being heard by a judge.

When the chance of “guilty” is consistently high (i.e. the vast majority of the money in the market is betting on guilty and there’s low volatility), the accused is moved to a long-term prison. Time spent incarcerated is treated as part of the sentence regardless of whether it’s before or after a judge looks at the case, meaning that if it takes 20 years for a judge to examine a case, but then the judge sentences the accused to 25 years in prison, they’ll only be kept for 5 more years. When a person’s case has been pending for longer than the expected sentence on the conditional market, they are typically released and treated as having served their time, with their case marked as closed. In this way, most people accused of a crime in Utopia never have their cases heard by a judge at all — market traders (often including the accused) are so confident in predicting a guilty verdict that the person simply sits in prison for a while and is then released when they’ve served the expected sentence.

For those cases that do reach a judge and result in a guilty verdict, a new prediction market is immediately created asking whether the verdict will be overturned if reviewed. These post-judgment markets serve as Utopia's version of an appeals system. However, due to a strong culture of judicial precedent and deference, these markets typically see minimal trading activity and usually predict that the verdict will stand. The exception occurs when traders identify clear judicial error or when genuinely new evidence emerges — in such cases, significant money flows into betting on reversal, triggering judicial review. This market-based threshold ensures that judges aren't overwhelmed with frivolous appeals while remaining responsive to cases where new evidence comes to light.

When the police and prosecution are clearly wrong, and market traders confidently predict a “not guilty” verdict, cases are often quickly dropped with profuse apologies. Why? Because in Utopia, those found not guilty are immediately and automatically entitled to compensation commensurate to the damages inflicted on them by law enforcement. For instance, if a doctor is wrongfully accused of assault and then serves three years in prison before being found not guilty, they’d be entitled to three years of lost earnings, plus the reputational damage from the wrongful arrest, plus the basic inconvenience and emotional damages from being imprisoned. Not-guilty findings are thus extremely costly for the government, and an immense amount of pressure is put on prosecutors and police to avoid arresting innocent people, especially for long periods of time. It is typical for prosecutors to make deals with those who face decent odds of being found not guilty to drop their case with a public apology in exchange for the accused waiving some or all of their compensation. Even when someone wants compensation and won’t agree to such a deal, they’re usually released, just to limit the damages.

Only when things aren’t clear — when some group of traders sees guilt and another comparable group sees innocence — is it truly vital that judgment be reached. The cases that judges actually hear are selected pseudo-randomly in a way that simultaneously prioritizes cases where there is a reasonable chance of innocence and a lot of money has been invested.16 Money may not be the best metric for which cases are important, but it’s one that keeps people honest. Utopia has professional speculators who search for cases that are weaker than their market probabilities make them seem, and invest in non-guilty verdicts not out of a particular sense of altruism, but rather because they think betting on such people is a good investment. Defendants who can expect significant compensation if they’re found not-guilty (e.g. those who have been in prison for a long time) will sometimes offer a fraction of their damages as a prize to people who bet on them in the guilt market in order to attract such speculators.

To reduce risk of having to pay large compensations, law enforcement usually doesn’t keep people in jail while they’re awaiting trial unless they’re judged by the prosecution to be likely to commit more crimes or flee into hiding. For instance, in most contract disputes, the accused party is never imprisoned. This lack of prison still holds even when someone is found guilty. All sentences are framed in terms of time in prison, but the government allows most criminals to pay off their time in a way that scales with the financial resources of the criminal, so they remain a fair and significant punishment. This helps keep members of society productive and reduces the financial burden of maintaining prisons. Repeat offenders and those otherwise believed to be an immediate threat are the major exception, where the need for incapacitation leads the justice system to keep them imprisoned without ability to buy freedom. In borderline cases, where the state is unsure of the danger, surveillance devices like ankle monitors are occasionally offered as a (voluntary) compromise.

Because all crimes are punished in a homogenous way, the distinction between “murder” and “manslaughter” or other criminal ontologies isn’t considered important in Utopia. Judges simply inflict longer sentences on those who are more egregiously at fault, or who are a higher risk to society. The distribution of sentence lengths (both given by judges and by the markets that emulate the judges) tends to be bi-modal, with short sentences for those who are at a low risk of re-offending,17 and long sentences for those who show consistent criminal behavior, even if that behavior is as mundane as shoplifting or violating contracts.

Utopia doesn’t really have a notion of “lawyer” as a profession. There are law-experts, of course, who teach in schools, write books, serve as prosecutors, and occasionally become judges. But many times those working on a case in law enforcement or helping someone who’s been accused are professionals with no particular expertise in law, and instead are good at tasks like collecting relevant testimony and evidence into a convincing narrative. It is up to the judge (and somewhat to wardens) to evaluate whether the information they’re getting is pertinent and high-quality. In practice, most of those who decide someone’s fate are the many speculators across the world, each trying to find an edge in the marketplace of justice.

Prerequisites: World Government
Foundational to: Part 2

Justice is sprawling and complex. Pulling on a single thread naturally brings the broader picture into question. In my recent essays on policing, we began tugging on that thread. Now, we follow it to the next logical stage: adjudication — the phase between when an arrest has been made or a suit has been filed and when a bargain or verdict is reached. In many ways this is heart of the legal system, and it deserves a careful look to ensure all concerns — philosophical or pragmatic — are being addressed and balanced well.

Criminal vs Civil Justice

Here I’m largely going to be focusing on criminal justice, which assumes there is a suspect who is arrested because of a specific law being broken, and that the government is the primary party pushing for justice. But before getting into the details, I want to spend a moment connecting criminal justice with civil justice. In my eyes the distinction between the two is closer to an implementation detail than a difference in kind.

To see the connection, let’s first imagine a world where there is no strong government, and all justice is the product of individuals or groups suing others in a “civil” system of community adjudication. If someone is assaulted, they can sue the attacker for the damages incurred. If someone is murdered, their family might sue for damages. If the community is fair and strong enough to hold the wrongdoer accountable, no government prosecutor or distinction between “private disputes” and “crimes” needs to exist.

A natural objection to this kind of system is that it might let some villains off too easily. Consider someone who goes around murdering orphans who have no family to seek justice on their behalf. Or consider a skilled thief who simply buys insurance that pays out if he’s caught.

But also, as I discussed in my essays on speed limits and surveillance, there is an additional set of victims to these actions besides those immediately harmed: those who were benefiting from living in a safe society. If I might be attacked while walking down the street, this hurts my well-being even if I’m never actually attacked. Thus, if someone else is attacked, I believe that it would be fair to sue the attacker for the harm to my sense of safety. While the damage to any one bystander is low, there is something akin to a collective social damage to violent acts. Through this lens we can almost see criminal law as akin to a class-action lawsuit.

We can likewise imagine an opposite setup, where there is no civil legal system, and the government is the only party that can initiate a court hearing. Things like breaches of contract, car accidents, and copyright disputes could, in this world, only be handled by the state taking action. But… presumably they could still take action! In order for contracts to mean anything in this world, it would need to be a crime to break a contract, but presumably one could simply set up the law this way, and have an aggrieved party report to the government when a contract was violated.

Consider a mega-corporation that routinely steals the ideas and works of small artists in an unfair way. In our world, each of these artists might be able to sue the company, and they might even be able to form a class-action lawsuit. But taking legal action is hard, scary, and risky; most artists might not have the resources to stand up for themselves, and the company might pay settlements to the few who do, or use their power and influence to pressure lawyers not to help the artists. In the world where there are only criminal cases, the artists might be free to request help from the state and then leave it in the hands of the government to ensure justice is done.

There is, of course, the question of whether the criminal justice system is efficient and effective. Civil suits might be better able to bring in specialist lawyers and make speedy deals, improving the quality of both process and outcomes. But again, that is an implementation detail. Regardless of whether framed as criminal or civil, justice is justice.

Utopian Universal Justice

In my conception of Utopia, different laws are enforced at different levels of government. Local communities of various sizes may have specific laws, norms, and rules that they enforce according to various standards. Some communities may rely on arbitration in private courts. Some may use a religious guide, condemning people for less-than-holy lifestyles. But in all such cases, emigration out of these communities remains a fundamental right, including for those found guilty of violating local laws. Someone who lives in a region with a local government may always choose to leave that region, even if it means avoiding punishment for their transgression.

Three factors help mitigate this loophole, though the third requires us to understand how the broader Utopian legal system works:

First, banishment and public shaming are common punishments in Utopia, as they’re incapable of being dodged through emigration.

Second, emigration requires a destination, and many regions run background checks and block immigration when the migrant is trying to dodge the consequences of having broken the laws of their previous country. There are usually some places willing to take such people, particularly if their “crime” was some silly, cultural taboo, but options are often limited.

Now, in contrast to local laws, a universal law is one that holds everywhere in the world. Murder, slavery, theft, and assault,1 for example, are always illegal. In addition to these (and other) basic crimes,2 I believe that Utopia treats breach of contract as a universal crime. If two parties mutually agree to some deal, it is in the interest of all of society, not just the involved parties, for that deal to be honored. Those involved are encouraged to settle the issue out of court, often through a private arbiter, in the interests of efficiency or to avoid jail time. But if no settlement can be reached, the world government will probably get involved. Most crimes processed at the world-level are contract violations.

This brings us to our third mitigating factor: Residents of a city or region are often required to sign explicit social contracts holding them to the laws of that place in order to enter or stay past a certain age. The universal law requires those who sign contracts to genuinely understand the deal on offer (and be uncoerced) in order for the contract to be (criminally) binding. This means children (and non-human animals) are generally held to fewer standards.

Because the laws of the world government apply to everyone, they must be both uncontroversial and simple. The Futarchy puts a significant weight on having a legal code that fits in a single book3 capable of being read in a single day by a precocious 13-year-old. This is partially accomplished through federalism—leaving all laws that could be handled by local governments to the local level. But it’s also accomplished through public prizes for anyone who can rewrite the laws of the world in a simpler form4 (without losing fidelity).

Expert Judgment

Let’s start our investigation into the adjudication process itself with the question: who decides whether someone is guilty? In many countries, including the US and UK, serious crimes require a verdict to come from a jury. Others, such as the Netherlands and Israel, entirely rely on professional judges to decide cases. (And some, like France and Germany, use a hybrid system.) The main reason for having random citizens produce a verdict, as far as I can tell, is as a check on corruption and tyranny. Judges have a lot of power, and historically (as well as in less civilized5 parts of the world) judges have been prone to being tools of the state and/or organized crime at the expense of the people. By putting the burden of judgment on the shoulders of random citizens, we reduce the risk of external incentives poisoning the process, and do so in a transparent way that helps establish trust.6

But while resisting corruption is a laudable goal… have you met people? Especially those who have nothing better to do than sit on a jury, or are too incompetent to mention jury nullification or otherwise get out of it?7 For every well-informed, attentive, wise member of the jury, you’ll find a half-dozen people who, in an ideal world, would never be issuing life-changing judgment on a stranger. Involving laypeople may be a step towards democracy, but I fear it also leads to three steps away from technocracy/meritocracy. Juries are not just vulnerable to ignorance, but to prejudice, emotional manipulation, and plain inattentiveness. The biases introduced by the selection process alone should be worrisome.

And while we’re on the topic, I should also note that forcing twelve people from their normal lives for the duration of a (sometimes lengthy!) trial is expensive. It’s not extremely expensive—most trials are less than a week long and most highly productive people will dodge jury duty—and one could argue that it’s worth it. But it’s still a cost in the tens of thousands of dollars per trial that could go towards other things (e.g. prizes for local journalists covering the courts).

Corruption needs to be fought against. Judges should be those of the highest character, and unimpeachable in their devotion to ideals. But the way forward is to actively defend wise elites from outside influence, not to resort to demagoguery.

Utopian Judges

Local governments may use juries, priests, robots, or slot machines to determine the outcome of their trials, but I believe the Utopian world government relies entirely on professional judges8 to handle its universal laws. Many resources are spent attracting wise, disciplined, and honorable experts for this role, including hefty salaries and teams of elite support staff. But the selection process to be a judge is also extremely strict, and because they preside over the entire world, the number of judges per capita is extremely low (~1 judge per million people). As a result of these low numbers, at any given time there are many thousands of cases per judge.

This shortage is lamentable, but also intentional. The Futarchy government intentionally keeps the selection criteria strict most of the time to maintain a high standard. If the number of cases per judge gets too high, standards are sometimes relaxed and more judges are hired. The Futarchy also sets a “tempo” to the justice system, which limits the maximum rate at which a judge can handle cases, usually to a fairly small number like 20/year. This is meant to encourage slow, careful verdicts. Judges are free to take as long as they’d like with a case, but in the absence of a rate limit, the most hasty judges would be the ones deciding the most cases.

By virtue of being carefully selected and highly compensated professionals, judges can be held to an extreme level of impartiality and scrutiny. In my Utopia, upon being hired, judges go through a two-year sequestering process where they and their family move to isolated communities9 called “cloisters” and spend months studying law, philosophy, and other subjects to prepare for their upcoming responsibilities. Cloisters also often have small monasteries, schools, laboratories, and other timeless institutions where people can focus on truth and enlightenment, away from the hustle and bustle of the outside world. The core property of a cloister is that news from the outside is heavily restricted by an organization of judicial managers called “wardens.”

Those in a cloister who aren’t wardens (including judges) have no easy access to news or visitors, and wardens go through a vetting process almost as strict as that for judges in order to maintain the information firewall.10 The two-year sequestering applies not just to the judge, but to all who would come and live in a cloister, with families moving progressively inward until finally being allowed to freely interact with others in the heart of the community. The natural two-year lag time for news inside a cloister11 combined with the absence of visitors from outside means judges are very hard to influence, at least when it comes to specific cases. This isolation is considered somewhat sacred, and the local governments of cloisters often banish those who violate the spirit of their system.

Another factor ensuring exceptional impartiality is that cases are assigned to judges in a fairly random way. Judges usually specialize in either violent crime, property crime, or (most commonly) contract crime, but within these divisions the allocation of cases to judges is completely random. (Judges (and senior wardens) can also recuse themselves from (or veto) specific cases if they believe there happens to be a conflict of interest, in which case it’s randomly re-assigned.)

Yet another factor helping impartiality is that Utopia takes “justice is blind” very seriously. Wardens usually redact information (e.g. the names, faces, and sometimes the ethnicities of those involved) from the cases that judges hear, and judges typically never meet anyone directly involved in the case. Instead, each court case is effectively a database. A judge is given a huge collection of information all at once, gathered by law enforcement, the defendant, and other involved parties. This case data will almost always involve transcripts of testimonies, as well as physical evidence, emails, photographs, recordings, and whatever else the involved parties believe is relevant. Judges are encouraged to ask questions related to their cases; many of which are transmitted via warden to outside the cloister, producing follow-up testimony or domain expertise.

Judges are compensated at a flat rate and are given a fixed-length tenure of twelve years (not counting the two year sequestering), after which they are forced to retire, often moving into academia or living off their savings. A judge can, in theory, hear only a few cases and collect a hefty paycheck for mostly sitting on the beach all day, but the hiring process is so strict that these kind of parasites almost never actually get hired, and judges are instead selected for being hard-working and worldly enough to empathize with those involved in the cases they hear, even after more than a decade in seclusion.12 The rulings of each judge are public information, and the prevailing culture expects published explanations about how and why each judicial decision was made. Prediction markets about both how many cases a judge will hear and how many of their verdicts will be reversed upon appeal are used to determine who is a good fit for the job.

But how do trials actually happen, especially with so many cases per judge? In Part 2 we’ll explore various procedural desiderata and how to handle trials in practice.

Prerequisites: None (… but I recommend Public Goods Funding and Adverse Selection!)

I have mixed feelings around dating apps. On one hand, I met my lovely wife on Bumble, and I credit OkCupid for getting me involved with the Rationality community after I moved cross-country as a young man. These were two of the most important connections in my life, and I genuinely have dating apps to thank for them.

On the other hand…

Jokes aside, dating apps can be pretty awful. Adverse selection means that the vast majority of accounts are either inactive, bots, or scammers. If you’re lucky, you’ll get matched with people who are single for the very simple reason that they’re terrible at having healthy relationships. Finding real matches involves slogging through profiles, desperately hunting for those rare few people who are simultaneously (1) human beings that you would want to date and (2) people who would want to date you. Then you get to desperately try to keep them from slipping away while also not coming across as too desperate.

It’s kinda wild that dating apps work at all. They certainly weren’t how people used to meet, and from my anecdotal experience, people mostly seem to think the old ways were better. But the older methods of meeting people are dying out, and this is largely because dating apps are beating them. There’s a growing sense that it’s rude or creepy to pursue/seduce/approach people in public settings,1 often including classmates and coworkers. Unless someone has made it clear that they’re interested in finding a dating partner, they’re seen as off-limits by default. Dating apps offer a rare sense of clear consent to be approached.

On top of selling a step-by-step solution to loneliness, dating apps advertise abundance. They sell a message that, in the age of the internet, relationship partners are everywhere (yes, even in your area), and you’re just a few easy pulls of the slot machine away from winning big. But, like a slot machine, these experiences are designed more to engage than to satisfy. And so they absorb energies that might’ve been more efficiently spent elsewhere, often leading burnout and despair.

The End of Matchmaking

The current culture2 around dating is really, really weird from a historical perspective. The typical human experience, which is engrained in both our evolved psychology and our culture, is to live in a world that centers around the collective will of an extended family, clan, or tribe. In these contexts, picking a mate was far less a question of individual preference and more of one of communal negotiation. The very concept of “dating,” for example, is fairly modern, emerging in the late 1800s and and early 1900s as courtship rituals focusing on family approval gradually fell out of fashion.

The historic (and prehistoric) norm used to be that whether someone’s children found good mates and went on to have grandchildren was of unparalleled importance. Labor, relative to accumulated capital wealth, used to be extremely important and the size of a family was, in many respects, a direct indication of the wealth and power of its members. A family that failed to have children would grow poor due to (absent any social safety net) the declining ability of elders to hunt, plow fields, or defend against attackers. By contrast, a family with many children, in addition to having lots of available labor, could use strategic marriages to strengthen bonds, effectively growing the size of the social support network, making everyone involved more robust to random misfortunes.

This isn’t to say that individual feeling or even romantic love had no part in the old decision making process — merely that it was part of the negotiation between the families that were involved. When elders are powerful, arranged marriages become the norm, with extreme cases like royal families occasionally making arrangements while the bride and/or groom are young children. We can understand the trajectory over the last two centuries as the converse: when family elders lose power (and material interest), they become increasingly passive spectators of the mating decisions of their kin.

These changes ultimately stem from the technological transformation of the industrial revolution, but they manifest in dozens of ways, direct and indirect. Urbanization and geographic mobility exploded the space of economic opportunities while simultaneously pulling people away from the places they grew up. Rising value of machines made land and physical labor relatively less valuable,3 pushing towards smaller families with more emphasis on education and gender egalitarianism.4 As people became more educated and the world changed at an unprecedented rate, traditionalism and religiosity fell away, replaced by unfettered individualism, largely focused on personal joy and excitement. Contraception, especially “the pill” and increasing availability and intensity of pornography further decoupled sex from traditional relationship structures, amplifying the focus on individual satisfaction.

While some of these changes, like decreasing rates of teenage pregnancy, seem solidly good, others, like decreasing overall fertility are more mixed, and have serious downsides. I see the end of community-based matchmaking as one such costly change. While arranged marriages can be oppressive and awful, the general participation of grandparents and other members of the extended family in the project of securing a mate seems broadly good. A well-aligned family can use their myriad social connections, wisdom, and yes, even social pressure, to help set up marriages that provide enduring benefit to both the couple and the community.

But going back to having significant family involvement in picking mates isn’t really an option any more than we can un-invent the steam engine. People with power and freedom will chafe and balk when pressured to make major life decisions by others, even close family. The only way out is through, and that means finding technologies and systems that can successfully fill the void left by the collapse of communal matchmaking, not just parasitically feed off the growing ache of loneliness.

A Recipe for Love

One of the most frustrating aspects of the current situation is that it’s not like getting humans to fall in love with each other is some impossible challenge. People mostly want to find long-term partners and form deep connections. And when just a few key ingredients are present, they will. But with one major exception, companies and other organizations either don’t have the power to invoke the real recipe for love, or they have no incentive to do so.

What do humans need to become infatuated, fall in love, and choose a mate?

1. Propinquity — People need to actually spend a bunch of time around each other. Physical proximity is excellent, and in general the higher the bandwidth the better, but there are many forms of this. Reading someone’s words can be sufficient (especially for nerds) but it usually takes a lot of words. Even in a physical setting it usually takes repeated exposure. “Love at first sight” is mostly a myth.

2. Homophily — People need to identify aspects of the other that resonate with their own situation. The most central aspect of this is having shared relationship goals, but being of a similar social status, economic class, ethnicity, worldview, education, religion, or merely having the same interests is also important. Crucially, these similarities need to be emphasized in a positive way during interaction, without focus on the gaps or disparities.

3. Validity — People need to see the other person as trustworthy and a valid potential partner in good social standing, such that they feel confident that they won’t be harmed by letting their guard down, either in terms of physical violence, emotional abuse, or reputational damage. The most common form of this is in “social proof,” where a community or friend group implicitly vouches for its members, but it can also stem from more abstract reputation, as long as it feels real to the individual.

4. Scarcity — Perhaps most importantly, people need feel like they can’t sit back and contemplate their options forever. Absent some kind of stakes or time pressure, people often become indecisive, picky, and superficial, still wanting to perhaps go on dates or have sex, but refusing to seriously invest or commit. This pressure can be general, such as when someone feels the ticking of their “biological clock,” or from a more specific fear, such as the idea that a particular person is “the one” and they’ll slip away unless action is taken.

Traditional settings effectively provided each of these four pillars. And, alas, dating apps offer almost none. (One could argue that they’re decent at homophily, but even there they’re not ideal.) Thankfully, there still is one major institution that helps millions5 of people find love each year: college.6

1. Colleges provide propinquity in the form of in-person classroom settings where students naturally spend time around each other day after day, as well as at parties and other locations on and around campus.

2. Colleges naturally produce strong homophily by being selective and placing students in the same context. Just the fact that two people are going to the same school is a good sign on priors for them having similar goals, education, status, and worldview.

3. Colleges provide easy validity by mixing potential lovebirds into a complex social scene complete with friend groups and reputational gossip.

4. And colleges provide scarcity in that everyone knows the clock is ticking down to graduation. This natural timeline creates a pressure to not only start dating but to seriously invest and form a bond that’s strong enough to survive the people involved leaving school, potentially traveling across the world, and finding work.

Indeed, in some circles there’s something of an open secret that one of the most important functions of elite universities is helping young people from upper and upper-middle class families find spouses (and network in general).

Unfortunately, this probably isn’t a scalable solution. Going to college is expensive and impractical for people who already have established jobs and careers. Moreover, the very timeline that makes college effective for matchmaking—late teens to early twenties—increasingly misaligns with modern life trajectories. Many people now spend their twenties focused on career development, experimentation, and self actualization. It’s increasingly common for people to only decide they’re ready to settle down as they approach and enter their thirties, sometimes long after their college years have passed. Yet if college demonstrates that the recipe for love still works when properly implemented, perhaps we should ask: What would it take to recreate these four essential ingredients—propinquity, homophily, validity, and scarcity—in formats that serve people wherever they are in life? How can we solidly fill the matchmaking void left by both the decline of traditional family involvement and the inadequacy of dating apps?

Signaling Singleness

One very basic starting point in moving to a better world is to bring back one of the primary uses of the wedding ring. Specifically, it used to be the case that if someone (especially someone young) didn’t have a wedding ring, they were assumed to be available as a potential mate. This is an extremely useful signal in that it creates social affordance for attempting to court/seduce someone. It also is part of a weird, toxic assumption that all unmarried people (especially women) are “fair game.”

The solution is easy: create an opt-in(!) common-knowledge signal that someone is hoping to find a new partner. These kinds of signals are arbitrary, but in one possible Utopia, I can imagine that people wear a single earring to indicate that they’re interested in finding a new partner. (Because they’re single… get it?)7 Perhaps a dangling earring could indicate a preference for men, and a stud-earring could indicate a preference for women.8

The diversity and complexity of jewelry means that it’s possible for many other signals to be sent in a similar way, such as preference for long-term relationships over more casual hook-ups, or whether someone does or doesn’t want kids. But I think it’s probably mostly a trap to try and encode many signals. People mostly have complex feelings around what they’re looking for, and those kinds of complexities are probably best left to actual conversations between people. Indeed, many people (especially young men) realize their preferences are actually quite flexible only after meeting someone who they really like.

The complexity of things is why I don’t suggest using left/right ear to indicate preference for men/women. Indeed, almost nothing is a true binary, and thus would be a mistake to associate with which side the earring is worn. But, because people will be people, culture will probably try to find something to indicate by left vs right, and thus I think it should be used for the one notable exception to the “just have a conversation about it” heuristic. I think Utopians wear their earring on their right ear to indicate a preference for flirting, and their left ear indicates a preference for more up-front, direct expressions of interest. Some people strongly dislike bluntness, and prefer to be slowly seduced, while others tend towards obliviousness and need a more direct approach. Importantly, unlike other relationship preferences, this one is best established before the parties have any conversation, and needs to be unambiguous, since it sets the scope of what kind of conversation is good.

A great thing about using jewelry in particular is that it serves as something of a fashion statement, serving its natural purpose of attracting the eye, and also allows the person to show off who they are. See someone with a black-skull stud earring with ruby eyes? That can tell you a lot about who they are! And it serves as an obvious (and flirty) conversation starter: “I love your earring!”

The Utopian earring signal is not a replacement for the wedding ring. The unending band of marriage is more significant than just “not looking for a new partner.” Plus, some people, such as those in polyamorous relationships, can wear both an earring and their wedding ring to help make it clear to people that they’re in an open relationship.

My guess is that Utopia also has at least one holiday oriented around the art of flirting, in which events (especially dances) are thrown for people looking for love where it’s understood that flirting with those wearing a single earring is actively good, rather than just allowed. Through holidays like this, Utopia helps teach young people the skills needed to find good mates and have fun while doing so.

Asking someone out who isn’t wearing an earring is not considered inherently rude in Utopia, but it is considered risky. If the person being asked out gives a harsh rejection, that’s allowed — they didn’t ask for that sort of attention — and further advances are seen as sexual harassment. But by contrast, it is considered a faux pas in Utopia for someone wearing an earring to be mean to someone who expresses interest in them, even if they’re not feeling it from that suitor. Making space for people to find the courage to ask someone out is the whole point of the earring signal, and while a polite refusal is always okay, a nasty put-down is the sort of thing that damages that space and hurts society.9

In Utopia, when someone asks someone else out on a date, it is strongly expected that the asker will pay for food, transportation, and everything else.

Utopian Mate Selection

People finding quality mates is an important problem, and Utopia doesn’t just settle for having a passive signal that someone is open to being approached. Utopia also has markets and infrastructure!

In particular, Utopia aligns the incentives of dating apps and similar services with those of their customers through a system of partnership bounties. When someone starts actively looking for a partner, they pay in any amount of money to an escrow service to serve as a bounty.10 Anyone can, with consent, see how big a person’s bounty is. Most dating services, such as dating apps, don’t charge any direct fee, but require a users to have a non-trivial bounty posted in order to participate.

Bounties are indefinite, and can’t be withdrawn.11 When two Utopians start dating, and go through various stages of the relationship escalator, one of the important steps to demonstrating commitment is for each of them to award their partnership bounties to whatever matchmaking service was most instrumental in helping bring them together.12 Thus, in order for a dating app to make money, it needs to get its users to seriously commit to each other and successfully exit the dating pool, aligning incentives.13 There’s a meme in Utopia that if you’re failing to find love, you should offer matchmakers more money.

Utopia has some online dating services/apps, and thanks to concepts like spam penalties, and the alignment of incentives, they’re slightly better than in our world. But, because they lack the ingredients of propinquity, validity, and scarcity, they tend to be outcompeted by matchmaking businesses that focus on more effective pathways to serious connection. In particular, the primary way that Utopians find mates is by going to singles events and venues, such as dances, often with assistance from a professional matchmaker who collaborates with venue hosts.

To concretely highlight one possibly story, let’s consider a young man named Adam who decides he wants to find a lady to partner up with, and maybe marry. He starts by saving up a day’s worth of income (e.g. ~$200) and posting it as a partnership bounty. He then creates an account on a popular dating billboard website, which verifies that he has a meaningful bounty, and writes a little bit about himself and what he’s looking for. A skilled grandmother named Beth who works part-time as a matchmaker, and pays a small fee to be able to search the website, spots his profile and sends him a message suggesting they make time for a video-call where she can give him tips on finding a partner in his area.

On the call, Beth tries to get a handle on who Adam is, and who to match him with. She finds out that he likes kayaking, and knows a girl his age named Carol who does group kayaking. She gives him some tips on flirting and suggests that they schedule a follow-up call so they can talk about how it went.

A week later, Adam tells Beth that he didn’t really like Carol because he didn’t think she was beautiful enough. Beth, wise and experienced, listens patiently, asks him about his sense of beauty, and nudges him towards being less superficial. Still, she accepts that maybe he needs to go on more dates and really understand the best choice problem. She says she has more ideas for beautiful women to set him up with, but she only works long-term with people who post a bounty of at least five times what he currently has. He grudgingly saves up, and then continues to receive her help.

Eventually, Adam meets Eve at a night-club and they fall in love. Even though Beth didn’t specifically tell Adam to go to that club, he credits her advice for helping him learn how to approach, flirt with, and ask people out, and so he awards his bounty to her as a way of expressing gratitude. Eve doesn’t share Adam’s perspective, and instead gives her bounty to the club. They break up the following year, and Adam begins his search again by posting a new bounty.

The search for lasting connection is still difficult in Utopia, even with the support of more people and institutions with aligned incentives. Dating is naturally hard. But with the right culture, institutions, and attitude, it can also be satisfying and magical.

Prerequisites: It’ll be helpful to have read Numbers and the first part of Spacetime Distances. The Onion Test for Scientific Explanations is also relevant. Nothing is strictly necessary.

Curved surfaces like spheres or the Earth can't be accurately described by a flat map. All maps of the Earth are forced to distort their subject matter in some way, but some distortions are better than others, depending on how you're using the map. The Mercator distortion is garbage for comparing sizes, the Gall-Peters distortion is garbage for understanding shapes, and even the Waterman Butterfly distorts space slightly near the edges of the segments, such as around Pakistan:

Even local maps are distorted, though it becomes less and less of an issue at smaller scales, when the curvature of the Earth isn't particularly noticeable. It seems like a worth-while custom to include information about which distortion a map uses, alongside the date that the map was made.

North Isn't Up

I'm glad our world has settled on a standard orientation for maps. North being normally at the top of the map means anyone can orient to a map just by seeing how the words are oriented.

Unfortunately, we chose the wrong standard.
The top of the average map should be east, like it used to be.

Imagine being a stationary god looking down on the Earth from space. The Earth is spinning under you, with the east retreating and the west arriving. Imagine there are words printed on it (for the moment consider the words being from a book). What word-orientation would be most convenient? Words are read left-to-right and top-to-bottom,1 so it would be nice if the Earth naturally rotated with the reading direction. This would put either south or east at the top. But south-up is a bit annoying because it's not clear how long each line of text should be, and you'll have way more words being half-visible or hard to read due to the curvature.

To orient literally means “to face east,” and how fitting it is to have the place where the day begins be where the eye begins to read. The sun rises in the east, and it seems fitting that the east should therefore be the part of the map that rises over the rest. When looking down on a globe it's convenient not to have a pole in the way (hence part of why most globes are tilted). Because the poles are inhospitable, long journeys are most commonly east-west, which (half of the time) matches how one faces when orienting with an east-up map. Last (and least) it'd be kinda nice if world maps put the most populous place/biggest continent (Asia) in a prominent position.

Truly, this seems overdetermined.2

Coordinates

What's your current latitude and longitude?

Most readers, I suspect, won't have a good answer. Even those with enough knowledge to guess at a rough number probably don't know their position with any real precision.

This seems a bit silly. It's not like it's unimportant to communicate precisely about where we, and other things, are! But instead of just saying plainly where something/someone is, we use a bizarre collection of local schema involving cities, streets, blocks, and postal codes, which are inconsistent between countries and don't even allow for talking about arbitrary locations.

Things didn't have to be this way. Imagine if all addresses were coordinates (as they effectively are in some places). You'd be able to easily navigate to an arbitrary destination without a map! Extremely convenient.

But latitudes and longitudes kinda suck.

First of all, they seem comparable, but are decidedly different indicators. If someone is at 43° latitude they're somewhere in a ring, centered on the North Pole. If someone is 43° longitude they're somewhere in a wedge, with tips at the poles, and centered on the west coast of Somalia near Jamame. There are 360 degrees of longitude, but only 180 degrees of latitude. Due to the fact that Earth isn’t perfectly spherical, one degree of latitude change does vary ever-so-slightly by location, but is consistently about 111 km, but moving a degree of longitude means very different things depending on how close you are to the poles.

Another problem with using latitudes and longitudes is their bulk. Consider the common use-case of indicating a postal address. I randomly looked up two townhouses in Manchester. Their longitudes were essentially identical, and their latitudes were very similar: 53.446635 versus 53.446603. Almost all of this information is saying that they’re in (the latitude of) Manchester, which is practically useless for locating the particular address. This problem becomes worse when also considering the longitude, which may or may not change along a given street.

Numbers like these are thoroughly unmemorable. Humans are excellent at remembering words, and can pack in far more information with a lexical index. It's this feature that's led to clever technologies like what3words, which would be great… except that involves dropping the ability to use addresses to navigate and reason about the world — an extremely useful feature!

Using numerical coordinates to refer to places on Earth seems good, but I claim that a more ideal system would discard latitude and simply count distance to the North Pole in kilometers, rather than angular distance from the equator. This would help highlight how it’s a different kind of beast than longitude.

Longitude is mostly okay. I think it should probably go from 0° to 360° instead of -180° to +180°, and if east is up, then it seems better for it to increase as it goes westward, rather than decrease, like it currently does, but these are nitpicks. It’s a bit of an accident3 that the antimeridian happens to be on the opposite side of the Earth to London, but I think it’s well positioned — the middle of the Pacific Ocean, being so empty from a human perspective, is the natural place to cut a map or start measuring from. I do think that it would be neat if we used hues to represent longitude, since humans naturally understand that color goes in a wheel. We could then reference a place with something like “5483 vivid-indigo4 kilometers,” rather than by using a pair of angles.

Utopian Addresses

The Utopian writing/number system, there’s a correspondence between small numbers and basic consonants:

0 = n/ng
1 = m
2 = k/g
3 = t/d
4 = p/b
5 = f/v/th

Using this correspondence, we can map two-syllable words into numbers from 0 to 215, by treating the consonants as digits in little-endian base-six.

 # | Utopian | Digits | English
----------------------------------
 0 | naunung |  0,0,0 | value
 1 |  mining |  1,0,0 | curiosity
 2 |  kanung |  2,0,0 | learn
 3 |  tinung |  3,0,0 | soul
 4 | paunaun |  4,0,0 | hint
 5 |  funung |  5,0,0 | culture
 6 |  neimin |  0,1,0 | yearning
 7 |   mamin |  1,1,0 | improve
 8 |  kimine |  2,1,0 | awaken
 9 |  tumine |  3,1,0 | inventor
10 |  pimine |  4,1,0 | newborn
11 | feimine |  5,1,0 | fusion
12 |   nikan |  0,2,0 | optimal
...

We could pack more information into our choice of words, such as by using vowels, additional consonants, or the structures of the words, but with just 216 options we can prioritize sensible words that sound reasonably different, which is helpful when communicating. (Imagine trying to distinguish what3words’ “///spin.lawyer.bath” vs “///spin.lawyer.path” over a radio or in a noisy pub.)

Small patches of Earth’s surface are very close to flat, and so for regional navigation it is most natural to reference the locations of things by reference to something like a Cartesian grid, approximately aligned with the cardinal directions. Ideally, within a given local area, addresses would be a pair of natural numbers, without needing negative or fractional values. Using the Utopian units of distance, we can say that a stretch (≈ 117 cm ≈ 46 inches) is too fine-grained for most purposes, while a tower (≈ 42 m ≈ 140 feet) is going to be too course for dense, urban places. The geometric mean, 6 stretches or 1/6 towers (≈ 7 m ≈ 23 feet), is just right.

216 of these units is exactly one Utopian mile (≈ 1.52 km), so while we could theoretically use a pair of words as an address, it would only work in the context of a neighborhood. If we instead use two words for east-west and two words for north-south, we can uniquely identify addresses within a province that’s about 329 km on each side — big enough to uniquely identify them with reasonable-length names. By using the province name in between the number-words, we can unambiguously locate any place on Earth in less than a dozen syllables.

For example, suppose there’s a province named “Olguin.” We can indicate an address there by saying:

tigab fikuk Olguin nikud pauneim

This looks like a pile of nonsense sounds, of course, since we don’t speak Utopian. But to someone for whom those are words, it might read something like:

Red net, Olguin, Crying motive.

Very memorable! It’s at least easy to say and memorize when compared to the latitude+longitude of “20.747797, -76.029233” or the plus code of “PXXC+387 Báguano, Cuba.” It also compares reasonably well to traditional address like “Union Street #7941, Báguano, Cuba” or a what3words like “///digesting.rookies.clicker.” And unlike those systems, it serves as a coordinate system that can be used for navigation and doesn’t require a database to use (except to know where Olguin is5):

tigab fikuk | nikud pauneim  (words)
t,g,b,f,k,k | n,k,d,p,n,m    (letters)
3,2,4,5,2,2 | 0,2,3,4,0,1    (heximal digits)
      19383 | 8760           (decimal numbers)
 136.534 km | 61.705 km      (distances)

Once they learn the system, Utopians can use this property of addresses to easily find places they’ve never been before (without the aid of machines).

Of course, not everything we want to reference when navigating will have an address. Towns, neighborhoods, mountains, and rivers get names like they do in our world, but what about roads? It seems right to me to give names to major roads for the sake of navigation. If all cities had roads set up in a grid, I’d suggest naming the roads after the shared part of the addresses they connect, plus some indicator of whether they’re north-south or east-west, but most roads aren’t so perfectly aligned (even in Utopia), and in the absence of that, I think it makes sense to fall back on monikers like “main street” and “broadway.”6

But it also seems unnecessary to give arbitrary names to all minor roads. At intersections, I think the Utopian norm is to see signs indicating the (numerical) address for that intersection, along with arrows pointing in the cardinal directions.7 Together, these bits of information are sufficient to help travelers navigate to their destinations.

One detail still remains: how, exactly, do we form a “grid” given that the Earth isn’t actually flat. On the small scale of a province it might not matter very much, but the principle still matters. My guess is that Utopia defaults to azimuthal equidistant projections (AEQD), just as our world does, for regional maps. Since addresses are based on distances, it seems better to use a projection that preserves distances (rather than something like areas), and AEQD has the advantage of being symmetric in its distortions, not requiring any arbitrary choices of reference points (beyond the center point of the province), and works nicely even when near the poles.

Provinces can overlap. These are geographic areas, not necessarily cultural/political units, and it seems fine to have neighboring countries (e.g. Belgium and Luxembourg) use different addresses even if they could theoretically share.

Myriahedral World Maps

While an AEQD projection works fine for a local region, it’s a bad choice when mapping the entire world.

The Waterman Butterfly looks much better for visualizing the whole world. Why? What makes it so good? I claim that it’s the interruptions — the sections of map where there’s a cut.

In a sense, the Butterfly is a bit like eight sub-maps stitched together. Because each sub-map is of a smaller region, the distortions from the projection are less egregious. Indeed, in the limit of infinite cuts, a map can be “perfectly conformal” and “perfectly area-preserving” — supposed limitations disappear! Except of course, that a map with that many cuts becomes confetti dust, and is basically useless. But the point remains that interruptions are under-explored, and can make maps nice. As we increase the number of cuts from just cutting along the antimeridian to cutting all over the place, we move from the realm of polyhedral projection to myriahedral projection.

Unless you happen to need to use the map for navigating,8 the primary downside to myriahedral projection is that it tends to put cuts in places we care about, such as in the middle of your home country. We could render each country separately, but then by default we’d lose all sense of borders and where countries are in relation to each other. The trick is to include redundancy — a Utopian myriahedral projection should have a low-distortion view on every place, but some places might be viewed multiple times. Once we have these two principles in hand: myriahedral interruptions and redundancy, we can find our way to a truly Utopian world map — a small atlas of local views, which together have the added benefit of conveying the shape of Earth:

Prerequisites: Fairness and Licensure are the two main Utopian Dreams essays that seem useful, though I reference a smattering of others in the Utopia section at the end. Basic financial literacy will also help a lot, especially regarding adverse selection.

Imagine you’re a software engineer at a promising tech startup, grabbing lunch in the company cafeteria. You overhear someone casually mention that a major acquisition deal, previously thought dead, is suddenly back on track and likely closing next week. Excited about the company's prospects, you use your brokerage app later that day to buy a thousand additional company shares, complementing what you’re getting as part of your standard compensation package.

But, whether you realized it or not, buying those shares was illegal. And not just parked-in-the-wrong-spot illegal, but a serious felony offense. “Trading on material, non-public information,” more commonly known as “insider trading,” can be punished with twenty years in federal prison and as much as five million dollars in fines (for individuals). And even when prosecutors decline to press criminal charges, the Securities and Exchange Commission (SEC) can still sue in civil court, demanding up to three times what the trade made in profit.

These risks aren’t just theory. In January of 2025, a Los Angeles engineer earned himself two years behind bars after making a half-million dollars off of a friend’s merger tip. And the year before, a man in Texas was charged for trading based on information that he overheard from his wife (without her knowledge), while she was taking work calls at home. Each year, the US federal government prosecutes approximately 36 cases like these.

At its core, the effort to prohibit insider trading is about encouraging investment and promoting fairness. If some people, especially wealthy executives, have access to more information than outsiders, they can exploit that advantage for profit. A strict libertarian might describe this as “selling the information” but most people see it as harmful exploitation.

The modern concept began to take shape in the United States following the market crash of 1929. While early common law cases like Strong v. Repide (1909)1 touched on the fiduciary duty of insiders, the Securities Exchange Act of 1934 was the landmark legislation. It created the Securities and Exchange Commission (SEC) and gave it broad authority to regulate securities markets, including prohibiting “manipulative and deceptive practices.” Initially, enforcement was relatively limited, but the reach of the SEC slowly grew, especially in the 1960s,2 to restrict and control a broad range of activities within the stock market.

Restrictions on insider trading are not specific to America, however. Every major financial market in the world has similar restrictions. The near-global consensus is that this behavior is clearly detrimental and needs to be illegal.3

Why Insider Trading is Banned

Why is trading on private information considered so harmful? The typical arguments usually include:

• Reduced Market Liquidity: If market makers (the entities facilitating trades) suspect they might often be trading against insiders with superior information, they face adverse selection. To protect themselves, they will widen the gap between the prices at which they buy and sell (the bid-ask spread). This makes trading more expensive4 and can reduce overall market liquidity (the quantity of available trades).

• Unfairness and Erosion of Trust: The perception that the market is rigged in favor of insiders discourages participation by ordinary investors. If outsiders feel they're constantly playing against loaded dice, they might take their capital elsewhere, such as speculating on land, thereby reducing investment in wealth-creating industry.

• Encouraging Manipulation: Allowing insiders to trade freely can create perverse incentives. Executives could potentially manipulate corporate events or disclosures to profit from subsequent stock price movements, rather than focusing on increasing long-term company value. They might even short their own company's stock before intentionally tanking a deal. Politicians and regulators could likewise trade based on confidential policy information.

These concerns are serious. A market rife with manipulation and exploitation will repel investors (and hurt those who are too foolish to stay away), ultimately stifling growth. Is it any wonder that basically every nation has fought to keep insider trading to a minimum?

Why Insider Trading Might Be Good

Considering the near-global ban on insider trading, it might be surprising to learn that not all economists are against it. The most notable of these was Henry Manne, whose 1966 book Insider Trading and the Stock Market argues that this kind of market regulation is ultimately wrongheaded. Let’s consider his ideas, and other arguments in favor of allowing insider trading:

• Enforcement is spotty, and leads to potential corruption.

  • It’s impossible to know exactly how much insider trading occurs, but it’s potentially quite a lot. Subtle forms surely happen all the time, where those who are able to see more than just public documents have small edges. But there are even cases that are somewhat blatant, such as the surge of trading (mostly shorting-selling and later buying options) that preceded President Trump’s tariff announcements at the beginning of April.

  • Time will tell whether any particular instance of insider trading gets punished, but the spotty enforcement and connection to the government means our current system is already heavily burdened by insider trading and opportunities for corruption. Any narrative that claims insider trading would ruin the market needs to account for the fact that the market continues to function despite the presence of some insider trading that we see currently.

• Deregulated trading means more accurate prices and faster updates.

  • Accurate prices are the foundation of efficient allocation of capital. If a company’s true value is different from its market price, such as due to a secret, investors may make foolish decisions as a result. An example of this might be a tech company hiding their losses on their new product in order to keep their stock price high, but as a result their competitors believe it to be more profitable than it is and waste effort replicating a dud.

  • Society usually tries to encourage accuracy and speed by mandating that companies publicly disclose important information (such as profits) without delay. These laws come with their own costs, such as potential competitive disadvantage with privately traded firms.

  • When insiders are allowed to trade, they predictably push stock prices towards their true values faster and more reliably than official press releases. For instance, as a business deal gets closer and closer to being made, the price of those stocks can gradually shift, instead of suddenly moving only after the deal goes through and is announced on the news.

• Day trading is mostly a waste. Market volatility is bad.

  • Even though accurate prices are important, I would argue that we, as a society, invest too much time and talent into tracking them. The significant majority of market activity (~85% perhaps) is uninformed trading based on little more than vibes, rumors, or fleeting sentiment — essentially “noise.”

  • Immense resources and intellectual talent are poured into predicting the foolish wanderings of uninformed traders. And while theoretically helpful for keeping the prices closer to their true values, this broad pattern is mostly a zero-sum game.

  • If there were stronger warnings against uninformed trading, such as the “unfair” playing field of insiders vs outsiders, there would probably be less volatility and thus fewer incentives for brilliant people to spend their careers trying to understand that randomness.

• Insider trading is a natural reward for whistleblowers.

  • Telling the public about the misconduct of your employer can be a very scary prospect. Whistleblowers can face severe retaliation, including job loss, demotion, costly legal battles, harassment, and industry blacklisting, making it difficult to find future employment. The risk of these substantial costs can understandably silence many who witness wrongdoing, even if they would otherwise feel a moral duty to speak up.

  • If an employee can legally short a company’s stock (and/or selling the tip to a wealthy capitalist) before blowing the whistle, they’d have a powerful financial incentive to do so. This incentive would both soften the natural costs on the whistleblower, and encourage them to speak up where they might otherwise see no reason to act.

• Contracts can still prevent abuse.

  • Legalizing insider trading doesn't mean normalizing all behavior. To prevent executives from deliberately trashing their firms, companies could ban certain forms of insider trading on a contractual level, similar to how IPOs are handled. Likewise, policy makers could be similarly bound by contract with their government not to exploit their position to manipulate policy for their own gain. The blunt tool of a blanket statutory ban could then be replaced by more nuanced, bespoke agreements.

• Laws should be crisp and clear, for a multitude of reasons.

  • The various rules and regulations around insider trading are notoriously complex and often rely on fuzzy concepts like “fiduciary duty” and discerning the exact source and materiality of information. This ambiguity creates room for corruption, can have a chilling effect on some forms of legal investment, pushes tax dollars into enforcement bureaucracies, and private dollars into legal teams. Societies where insider trading is legal are more able to operate on firm principles, such as judging people based on their actions, rather than their motivations.

These arguments should not be taken as a claim that legalizing insider trading isn’t without costs. The kinds of stock markets we have today are almost certainly dependent on such bans, and those stock markets have done a lot of good. But I think it’s also worth tracking the benefits that legalizing insider trading could bring, since there may be alternative avenues towards high levels of capital investment.

Revisiting the Problems

The three main issues with allowing insider trading are increased bid-ask spreads, fairness concerns, and risks of manipulation by executives. This last one can be handled by contract law, but what about the first two?

I claim that banning insider trading is a bizarre application of “fairness.” Imagine if “tall person basketball” was banned out of a sense that it’s unfair to short people that tall people have this natural advantage. I agree that in this world people below the height limit might sometimes have more fun playing basketball, but I wouldn’t describe that as any more “fair” than our world. Height would still (mostly) be an advantage, right up until it was disqualifying, and one could argue that it would be unfair to the people who would want to play the game but are too tall.

It seems reasonable for outsiders to demand prices that appear to favor them, when trading with insiders, and claim that otherwise it would be unfair, similar to how a short basketball player might demand a handicap. And indeed, this is exactly what it means for a market maker to widen the bid-ask spread when potentially dealing with an insider. Indeed, it even seems fine to ban insiders from trading on a particular market, akin to a particular basketball league that bans a class of people with an obvious (statistical) edge. But these are a far cry from a total ban.

So, if the fairness concern doesn’t hold up upon reflection, the only remaining problem with insider trading is that it pushes outsiders out of the market by widening bid-ask spreads. Worse prices for outsiders means fewer trades and less liquidity, which can lead to less overall investment in economic growth, damaging the greatest engine for making the world a better place.

This sounds pretty bad! But is it actually true that legalizing insider trading would result in less investment? Companies would still pay the same dividends, and be potentially juicy investments… If outsiders could somehow convince the market maker that they’re genuinely outsiders who don’t have privileged information, market makers could offer them discounted prices (i.e. narrowing the bid-ask spread), due to not having to guard against adverse selection.

And indeed, we see this all the time! Many market makers will offer price improvements to traders who they suspect are making uninformed, long-term investments in the market, such as individual retail investors saving for retirement. Since market makers are in competition with each other, there is a natural incentive to snatch up uninformed investors by offering them better prices.

Utopian Stock Markets

Investing in the future is how the world grows rich, and the most natural form of investment is in building companies.5 Utopia heavily prioritizes corporate investment, especially in contrast to consumption and speculation. Land value taxes make land speculation naturally unprofitable, and steeply progressive consumption taxes mean that very high earners are incentivized to invest almost all of their money, putting it towards productive uses, instead of buying luxury goods like boats. Even those on the poorer end of Utopia will often invest some money, as holding growing assets is widely understood to be a good way to save, rather than holding cash or physical goods.6

At the same time that Utopia encourages investment, it strongly discourages random people from deciding how to invest. In addition to the common wisdom that average people can’t beat the market,7 many local governments make it illegal for citizens to trade individual stocks without getting a license and proving basic financial literacy, partly because it is normal for insider trading to be totally legal in Utopia. Instead of picking stocks directly, almost all investors buy into funds.

An index fund automatically trades to balance a diversified portfolio based on straightforward metrics like market valuation, and such funds are the norm in Utopia. Because these funds are diverse, and execute trades according to a fixed, visible algorithm, market makers know they’re at low risk of being exploited when trading with those funds, and offer them prices that are comparable to our world.

But Utopia also has many funds that are managed by professionals who specialize in high-information trading within a specific sector of the economy. These managers work to cultivate relationships and insider status within companies, helping bring investment capital to bear in exchange for getting access to financial data, being able to sit in on important meetings, and other fundamentals. Most “insider trading” happens between these fund managers as they effectively make bets against each other based on their models of company trajectories and industry trends.

Executives and other decision makers inside Utopian companies are, as one might expect, bound by contract to not use their power to hurt the company for their own financial gain. This is typically accomplished by limiting when executives can sell company stock, and prohibiting them from short-selling (or otherwise betting against) the company until several months after they leave. While laws and other policies in Utopia are usually set by futarchy, and thus robust to similar manipulation, in the instances where a government selects a person to set policy or otherwise be in an executive role, they are similarly bound by contracts with the state.

Outside of the paternalism of licenses, Utopia puts up very few barriers to trading stocks (or bonds or certificates), and generally has far fewer market regulations. All stocks can be shorted, for instance. Company shares can be bought/sold on many different stock markets, including in countries other than where those companies are based. Markets tend to be open to everyone with a license, and operate 24 hours a day. This competition between markets means that very few have trading curbs (“circuit breakers”), and Utopians generally see market movement, including rapid decreases in price, as reflecting a realignment to the truth.8 Caveat emptor.

All companies in Utopia have the same quantity of shares: one. When someone buys or sells a share of a company, the amount is always expressed as a fraction of the company, rather than in arbitrary countable units. The primary way this is talked about uses order-of-magnitude modifiers for units — we can imagine someone wanting to buy “250 micro-shares,” for example. Many markets put a limit on the precision of how granularly stock can be split during a trade (just for accounting/computing purposes), but in theory, fractional shares can be traded at arbitrary precision.

Prerequisites: Nothing strict, but reading Opinions will help understand prediction markets

Why do so many employees of American companies live in the USA? The answer is obvious: because that’s where the work needs to be done! Whether you’re a plumber or a store manager, you need to be at a specific, physical location to do your job, and commuting from Mexico every day is a teensy-bit impractical.

But what about “knowledge work”? About 60% of the American labor force are knowledge workers — consulting, IT, administration, finance, design, media, legal, and similar professions have far less in the way of physical constraints tying them to a specific location. There are exceptions, of course, both in the sense of there being jobs that are tied to a specific location despite theoretically being knowledge work (e.g. scientists working with expensive equipment) and there being specific periods where physical presence is particularly important (e.g. architects getting a sense of a building site). But in general, most of these people could be living mainly in other places, and perhaps commuting into the USA when needed.

The cost of living in Mexico is about half of that of the USA. Other countries in similar time zones (Columbia, Ecuador, Nicaragua) can be even cheaper, and if one is free to live anywhere, it’s simple for someone to support a family of five in rural Vietnam or India, just on US minimum wage. Yes, there can be some additional drawbacks to living in a poorer place (e.g. less access to quality healthcare), but these can often be at least partly compensated for through travel.

The real reason it’s rare for knowledge workers to live in cheap places is complex, and involves a combination of bad infrastructure, cultural inertia, and bad policies.

Some of this changed during the pandemic. Companies that had previously insisted on in-office work suddenly became open to working from home and began exploring and adopting technologies that can make remote work comparably productive. Unfortunately, the pendulum now seems to be on the back-swing, and old cultures of in-office work are reasserting themselves, mostly for bad reasons. Remote work is decidedly not for every person or every job, but I believe careful analysis shows that it should more often be the default for intellectual occupations.

The Massive Benefits of Telecommuting

Before exploring the costs, drawbacks, and counter-pressures for remote work, let’s go through the reasons why it can be good.

• Most obviously, people who work remotely usually have no commute. And even in cases where they use a co-working space, they usually don’t have to travel as far as they would for an in-office job.

  • I’ve been talking about on this blog since the beginning: time wasted in transit is comparable to lives lost. Reducing commuting time on work days by just 20 minutes for 1% of the world1 is about 800,000 years worth of time saved per year — 10,000 lifetimes.

  • In addition to burning life via wasting time, driving is also one of the leading causes of dangerous accidents. A reduction of that size would prevent ~2,500 deaths each year, and hundreds of thousands of injuries.2

  • And more mundanely, less need to drive also means less stress, needing fewer cars, spending less on roads, and less congestion during rush hour. In addition to the local costs, this would also reduce pollution and be generally more environmentally friendly.

• At least as important as the ease of not having to travel, telecommuting allows workers the flexibility to arrange their work to match their needs (thereby often increasing productivity).

  • A typical example of this is that some people prefer being by themselves and being able to focus on a task without interruption, while others feel energized by an open workspace with lots of other workers visible. In a single office setting it’s hard to accommodate both preferences, but if each worker is doing their own thing, introverts can work from home by themselves, while extroverts can join co-working spaces optimized for having the right vibe.

  • Sometimes someone might have an infectious disease, such as a cold, but feel good enough to do work. Remote work allows that person to be productive where they otherwise couldn’t without risking coworker’s health.

  • Working from home is often vastly more convenient for caretakers, such as parents, who can often still accomplish a lot, but have highly specific needs. Women tend to benefit more from remote work arrangements than men, due to more often being in caretaking roles.

  • This increased flexibility is likewise invaluable for workers with disabilities or other special needs that extend beyond basic preferences. Getting to and from an office can be a huge obstacle for someone with mobility issues, as just one example.

• A somewhat distinct form of flexibility is the way remote work allows employers to more easily deploy workers as needed, including to provide 24/7 support or to handle remote locations.

• Beyond ease and flexibility, the primary benefit for workers is being able to live in more affordable places, where the cost of housing and other necessities is lower. (And save money on transportation costs.)

  • This benefit doesn’t just go to workers. Depending on negotiating power, employers can benefit from low costs-of-living by hiring workers who will happily work for lower wages.

• Employers benefit from being able to cast a wider net when searching for talent, increasing competitiveness for job postings. Likewise, remote work means increased opportunities for workers, especially people in poorer communities.

• Much like other trends in globalization, remote work is disproportionately good for talented, hard workers who live in less developed parts of the world. The flow of resources to these people and communities is egalitarian, and naturally reduces global inequality3 and extreme poverty.

• Workers also just generally like being able to work remotely. Surveys clearly show satisfaction for remote or hybrid jobs being higher, on average. And thanks to the flexibility of remote work, this satisfaction can translate directly into increased company loyalty and reduced rates of turnover.

These benefits — reduced commute, flexibility, affordability, competition, fairness, and satisfaction — are huge when taken together. So why wouldn’t companies and workers be eager to go remote?

Infrastructure Investment

Thirty years ago, remote work was rare.4 Home internet, when people even had it, was primarily dial-up. Many people didn’t have home computers. And the idea of holding an important meeting in a chat room would’ve been a joke. Yes, there were a few early-adopters who worked mostly from home, such as authors or early computer programmers (including my father), but certainly nobody who needed real-time interaction could move to Thailand and expect to keep their job.

This changed, bit by bit, decade by decade, until the Covid pandemic caused many businesses to suddenly take stock and realize that it was now largely possible to work from anywhere in the world — in theory. Meetings could be handled on Zoom. Documents could be worked on together on Google Docs. And more specific needs could usually be handled through a variety of less well-known pieces of software.

But despite an abundance of technologies which make remote work possible, there are still several physical/technological barriers in practice. Most notable are speed and reliability of internet connections and the transportation of goods and people.

While high-speed internet is commonplace, it is not ubiquitous. In particular, there’s a catch-22 where poor places — whether they are rural areas of countries like the USA or India, or merely typical areas of countries like Namibia and Cuba — do not have the digital infrastructure to support remote workers, which means these communities have neither the resources nor demand for building high-speed connections necessary to attract them. And even when a place theoretically has a fiber-optic connection or other high-speed service, it often matters a lot whether that connection can be relied upon every day. Nigeria, for instance, has plenty of broadband internet availability, but suffers from an extremely unstable power grid that gives out over six times a week for an average of 12 hours each time.

Physical transportation infrastructure, whether it’s for freight or humans, also matters. Some knowledge work, such as biology, can sometimes be done remotely as long as there’s high-speed (and possibly refrigerated) shipping. More common is for remote workers to need to regularly fly or drive into a city to handle particularly important meetings or similar events. In both of these cases, lack of high-speed infrastructure can make all the difference. And beyond the necessities of the job, there’s a natural cost to being far away from civilization, and workers may not want to go remote if it means losing quick access to deliveries or hospitals.

Culture and Trust

But even though it’s important, infrastructure is not the main impediment to remote work. Rather, I think the biggest factor is cultural and psychological. Specifically, employers have a hard time trusting employees, and in-person environments are central to how humans build and maintain trust.

I got my first full-time job (making software) because two small businesses were sharing an office. I went in to apply to a job posting for one of them, and the guy I was interviewing with said, approximately: “I’m not going to hire you. You’re too weird. But my friend might be interested.” I then had my first interaction with my soon-to-be boss face-to-face (which I think was probably vital, since my résumé at the time was hot garbage). Other work that I’ve found has been similarly connected to the people I’ve spent time with face-to-face, and I think my experience there is not atypical. Not only do I know several people with similar stories of finding work through personal connections of various kinds, but I also know multiple business owners who actively try to hire people who are in their local communities and have personal relationships with people they know.

Looking to hire people through local connections seems fine, but why would it be someone’s preference, especially in domains like software, where remote work is easy? By casting a bigger net, and being willing to hire people from anywhere, I think it’s pretty straightforward to show higher expected productivity per dollar.

But the statistically expected productivity per dollar is actually not what most employers care most about. A star employee can potentially bring millions of dollars of value over their career, but the wrong person getting hired can ruin everything.5 Even productive employees can do immense damage to the company as a whole by breaking the law, alienating customers/clients/partners, or creating headaches for management. But the biggest risks usually come from the effects that bad employees can have on other workers and on company culture. Humans are social creatures and rely a lot on cues from peers as to how to behave; a single person visibly slacking off or criticizing the company’s direction can cascade into widespread inefficiency and chaos.

In other words, employers are naturally risk-averse when it comes to hiring, and so they select for workers they can trust (especially to maintain morale and culture), as well as those who seem like they’ll be productive. By hiring people who have an established reputation and who seem, face-to-face like they’ll be solid and productive, companies can better guard themselves against accidentally hiring someone who will do a lot of damage.

Similar dynamics are in play when companies decide who to promote. Workers who are physically present in an office have more opportunities to demonstrate that they are positively contributing to company culture, especially with their peers, while remote workers often have fewer such opportunities or are simply overlooked.

As long as companies are looking for these kinds of personal connections when hiring and promoting, network effects will dominate. Young professionals looking to get ahead in their careers will want to live in cities with lots of companies in their field, and work in offices where they can meet lots of coworkers. (Those with more established careers and bargaining power are more often remote.) And since ambitious professionals live in the cities, companies will naturally set up workplaces there.

These network effects form a self-reinforcing cycle. Since the most ambitious people will be urban, if someone lives in the countryside or wants to work remotely, this is a sign of lack of ambition, further pressuring those who want to signal ambition to live in the city.

Lock-In Policies

One might wonder why companies don’t just fire more people. Why not simply check who is talented, ambitious, and alignment with leadership, and then get rid of those who don’t measure up? There are several reasons. First, it’s not always obvious to management who is boosting and who is corroding company culture, both because “company culture” is vague and hard to measure, and also because disaffected workers usually try to hide their misalignment (so they don’t get fired!). Even when such firings happen, it’s often too late — influenced coworkers can continue to work against the leadership’s direction, and morale can drop as remaining workers fear that they might also be fired.

But often, the risks go deeper, and are tied to legal issues. Many places, such as Western Europe, Japan, Brazil, and Mexico have strong employee protection laws, but even in the USA, which is somewhat infamous for allowing employers to fire people “at-will,” has a variety of protections in place for workers, including anti-discrimination laws and the need to buy unemployment insurance. Additional restrictions come into play for fields with strong unions, such as education, journalism, government, entertainment, and healthcare. Even when an employer is justified in firing someone, the legal costs of fending off a wrongful-termination lawsuit can be intimidating.

These barriers are part of why businesses sometimes engage in “silent layoffs” — making employees lives worse and worse until they quit of their own initiative. (Another big reason is that many employment contracts involve employees losing severance pay or other benefits if they decide to quit, rather than being fired.) It’s hard to know for sure in any particular case, but part of what’s driving recent trends in demanding that employees return to the office is exactly this kind of stealthy downsizing.

There are other legal policies at play in return-to-office mandates, as well. One major factor is that after Covid drove many businesses to go primarily remote (or hybrid), offices in many city centers plummeted in value.

As fewer people commuted into the urban core, businesses that depended on those workers are struggling, and city governments (which often rely heavily on property taxes) have seen sharp decreases in funds. Zoning laws and other regulations have meant it’s illegal to turn office space into apartments, and so cities have been using a mixture of tax incentives and fees to try to pressure companies to return to previous patterns of in-person work.

Utopian Remote Work

I think Utopia takes the need for high trust working environments very seriously. Low trust means having managers surveilling employees constantly, which is no fun for anyone, wastes resources, and imposes serious constraints, including reducing the feasibility of remote work. Part of how Utopia keeps trust high is cultural: people in Utopia put more emphasis on honor, honesty, and other virtues, and aren’t shy about holding each other to high standards. Another aspect is higher general levels of management competence6 and deliberate skill-building in communication. But the lion’s share of the problem is handled by three major pieces of infrastructure:

1. Prediction Markets

   1. When a person — let’s call her Alice — is looking for employment, she typically contracts a third party7 to create two prediction markets:

      1. “How long will Alice’s next full-time8 job last?”

      2. “How highly will Alice’s next employer rate her after she leaves?”

   2. These markets are used to signal to potential employers. But when a company — let’s call it McMart — is seriously considering hiring someone, it usually also creates a trio of more specific markets:

      1. “Conditional on McMart hiring Alice for job X in the next Y months, how long will she work at McMart?”

      2. “Conditional on …, will she leave on good terms?”

      3. “Conditional on …, will she be at the center of any scandals?”

2. Gossip Brokers

   1. While prediction markets are a good start, they’re not nearly detailed enough to tell an employer whether someone will be a good fit for their team. What would be nice is being able to interview people who know things about Alice’s character and professional history. In our world this is mostly done by providing references and work history on a résumé, but it’s often hard to follow up on these and there’s an obvious bias, since the applicant can choose whom to include.

   2. Utopia solves this problem by having companies who are in the business of buying and selling gossip. These “gossip brokers” pay nontrivial money to interview people about the people they know, especially coworkers or professional colleagues.9 The brokers then compile dossiers on people, available for purchase by potential employers. By default, sources are anonymized, but can be revealed for a significant fee, which then gets split with the source as compensation for being exposed.

   3. Gossip brokers are subject to a lot of scrutiny (by media, customers, and other brokers), and compete with each other on integrity.10 Nobody wants to buy lies!

   4. It’s generally considered unhealthy for people to check their own dossiers, or at least to do so too frequently. Instead there’s a common cultural story that the best way to know what people are saying about you is to behave in a deliberate, consistent way, and trust that they’re saying you’re that person.

3. Talent Camps

   1. Because face-to-face interaction is one of the best ways to build and maintain trust, in addition to the more market-based systems, Utopia has a plethora of venues where talented people (especially young people) are invited to come and network with each other and with scouts for potential employers.

   2. These usually take the form of an explicitly programmed “camp” experience where a batch of people looking for work show up at a venue alongside a batch of scouts and managers for various companies. The program usually goes for about a week, and involves shared meals, collaborative exercises, and competitions between teams. Many such camps have only a few administrators, and rely on volunteer labor provided by the participants.

When a person is brought on to a Utopian team, by default they are expected to work in the same space as others on that team for at least a couple weeks as they get on-boarded and learn their role. This often involves staying in people’s guest-rooms or at hotels when the team they’re joining is remote, and can involve spending different weeks with different members of the team.

For knowledge work, the Utopian default is for workers to be able to locate themselves freely across the world. Such teams often go on retreats together for a week or two each year to maintain organizational cohesion. Despite the cost of flights and temporary living/working space, these teams usually save money compared to entirely in-person teams, due to not needing to pay for office space or housing in high-demand areas.

Due to increased prevalence of remote work, Utopia has more widespread access to high-speed infrastructure, and remote workers are seen as the lifeblood of many rural communities. Likewise, the hearts of cities are made less dense due to needing fewer offices. Utopian businesses generally have more ability to downsize at will, which leads to increased ability to hire without fear and less insanity around silent layoffs.

Prerequisites: None

Zeno’s paradoxes are wild. It’s easy, with the benefit of knowledge and hindsight, to see them as trivial, but I claim that there’s something highly surprising about reality that these thought experiments highlight: smoothness.

Consider the worlds of basically all games, which are not smooth. This is most obvious in board games, where the space is divided into squares or tiles, but it’s also true in video games. Despite the appearance of smooth motion on the surface, all video games1 represent the positions and velocities of their entities using digital bits that have a maximum precision. This means that in a game, time is divided into discrete steps and space into discrete positions; at any given moment the characters effectively move a whole number of spaces on an extremely fine-grained grid. At the extremes, this can lead to unexpected behavior and glitching, especially when things collide.

By contrast the world is extremely smooth.2 Not only is it true that between any two distinct points in spacetime there is a distinct midpoint, but there are also uncountably many other points in that interval that can’t be reached through any process of arithmetic on integers.

We can see nature’s preference for smoothness over discreteness again and again in everything from the way objects gradually move from place to place, to the way that their velocity changes gradually, to the way that things heat up or cool down gradually, to the way that things can be cut and subdivided, to the way that there is a continuous path between any two shapes or images, to the way in which water can freeze and boil, to the mathematics of circles and other natural shapes. Despite not having access to slow-motion cameras or the internet, even the ancients had hundreds of examples of this pattern.

And yet, they mostly didn’t take it to heart. Leucippus and Kaṇāda independently postulated that the world was composed of eternal, indivisible particles — atoms. And more generally, most people held that there were distinct categories of things — fire as distinct from water, men as distinct from women, rocks as distinct from animals, and so on — and that these categories were true divisions in reality, rather than an imposed and imperfect framework. Even now, most people do not appreciate the principle of smoothness and how far it goes in characterizing the world. There are an uncountable number of mid-points between all true things, regardless of whether they’re genders or “fundamental particles.”

Humans Love Simple Patterns

To operate in the world we look for patterns and structure, especially simple ones. One of the simplest and most useful cognitive strategies is to carve reality up into discrete chunks and categories. “This is a cat. That is a dog. She is Mom. He is Dad. This is the first carrot. That is the second carrot. This stone is not a carrot. I won’t count it. There are two carrots, overall.” This kind of mental motion is so fundamental that it’s hard to even notice it, and I would never suggest giving it up — it’s simply too useful. But I think it’s worth noticing how, when thinking along these lines, we’re making things up. There isn’t actually an underlying, objective truth about how many carrots are present. “Carrot” is an invention that lives in the mind, and only imperfectly matches reality.

Sometimes this imperfect match is clearly good enough. Even if it’s true that there’s no fundamental distinction between carrots and rocks (e.g. they are both made of electrons and quarks), it is extremely rare for one to change into the other or to encounter something that’s halfway between, and so the distinction holds up for day-to-day use.

The simplicity of these kinds of ontologies is part of their utility. If I were to suggest that you should drop the notion of “carrot” as a category, and instead force yourself to constantly notice the distinct shape and character of each object as a unique configuration of matter… Well, I wouldn’t begrudge you from ignoring me from that point on. It is true that carrots exist, and that notion is useful, but, simultaneously, it is not perfectly true.

Searching for Continuity

The principle of smoothness is useful in two ways:

1. It helps avoid map-territory errors and points to the complexities of reality.

2. It can steer us towards more realistic theories, both by:

   1. asking what a mid-point between things would look like, and

   2. asking why we don’t see that mid-point in practice.

Consider the notion that cats and dogs are “different species” for example. I claim that an ancient philosopher who truly internalized the principle of smoothness would respond to this idea by saying: “I agree that we can, in practice, distinguish between cats and dogs. But I bet a part-dog-part-cat animal is theoretically possible, even if there’s some important reason we don’t see any such creatures walking around.” And that philosopher would have been right! They’d be extremely, prophetically right! Thousands of years before Darwin, they would be reaching for an origin of species that explains why cat-like creatures split from dog-like creatures.

Similarly, some people think that there is a distinction between living and unliving things. Our ancient philosopher would again be prophetic in guessing that there is a continuity of semi-alive things in between, and that there’s no fundamental or discrete difference between life and non-life. The same insight can lead to appreciating the ways in which humans are a kind of animal, and the ways in which animals can be more like people than is generally recognized. And, more speculatively, it suggests that some (potential) beings are half-conscious, for all notions of consciousness. It predicts there are things that only possess a fraction of a mind or a thought. It predicts it’s possible to be partially dead. If humans “have a soul,” then the principle claims that it is theoretically possible to have half a soul (or to have 1/√2 souls!).

Attending to why mid-points don’t show up in practice can be a particularly rewarding line of inquiry. Why aren’t more things half-solid and half-liquid? Because these kinds of materials tend to only be stable at a particular temperature or with specific substances. What does temperature have to do with things? Which substances? Why those substances, and not others? The philosopher might not know how to answer these questions, but they can see their ignorance more clearly, and know to be searching for answers.

“Fundamental” Physics

Even as we acknowledge the advantages to adopting the principle of smoothness, it should be noted that over-zealous application can be dangerous. If our hypothetical ancient philosopher had been in dialogue with Leucippus or Kaṇāda, for example, they likely would have argued that the atomic theory of matter is wrongheaded, and that at the fundamental level, there are no firm divisions between things. But of course the world is made of atoms! They would’ve been wrong! Or rather, they would’ve failed to predict the discovery of atoms.

But, in fact, it turns out there isn’t a firm division between different kinds of atoms. We now understand that everything from hydrogen to uranium actually can be split into more fundamental stuff, and there are examples of weird, intermediate forms of matter, just as the principle of smoothness indicates.

But perhaps those in favor of atoms and/or elemental building-blocks simply jumped the gun? After all, we now think of electrons and quarks as distinct, fundamental particles. That’s not very smooth! The sharp distinction between different particles is antithetical to smoothness, as is the very notion of a countable “particle” that exists at a specific point. Smoothness says we should expect there to be parts of space that contain “half a particle” or bits of matter that are part-electron and part-quark.

And it turns out, that if you get really deep into theoretical physics, this smoothness is quite likely correct! The notion of “a particle at a point” starts falling apart in quantum mechanics, which tells us that reality is actually a smooth continuity across “configurations.” But the configurations in quantum mechanics repeat the idea of distinct particles at specific locations, and we can correctly infer that this means that quantum mechanics isn’t fundamental. Moving deeper, the best theory of reality that we currently have is probably quantum field theory (QFT), which replaces the notion of discrete particles with smooth, continuous fields.3

I certainly don’t know what the ultimate nature of reality is, but I think the principle of smoothness can continue to guide us in finding even more fundamental theories. For instance, the field configurations in QFT are drawn from a Sobolev space, which (among other things) means they’re smooth. But just how smooth are they? This is an open question in physics. My guess is infinitely smooth! More speculatively, I would guess that “energy is quantized” not because intermediate energy levels are impossible,4 but rather that they’re naturally unstable and/or difficult to reach.

As I reflect on it, I can only think of one possible “exception” to the principle of smoothness — only one place where I see discrete integers in the fundamental fabric of reality and I don’t expect that’s a human limitation: the dimensionality of the universe. Whether we live in three dimensions of space and one dimension of time, or some eleven-dimensional alternative, I expect that it makes sense to to think of these dimensions as discrete and fixed.5

Utopian Smoothness

Scientific and philosophical education in Utopia centers around two complementary approaches. The first is seen in The Onion Test, which highlights that almost all knowledge is tentative and incomplete, and that there’s a deeper layer of understanding waiting to be had. The other is First Principles Reasoning, which seeks to find strong foundations for knowledge and understanding in the form of clean, elegant truths that can be applied over and over again. Utopians debate which principles are worth holding as central, but we can see some of the more important ones showing up as recurring themes on this blog.6

Smoothness is a principle that pulls a lot of philosophical weight, both in physics and elsewhere. It is usually taught as foundational to our view of reality, alongside determinism, relativity, and the relationship between space and time. Thanks to grappling with this idea from a young age, topics like irrational numbers, calculus, and advanced physics seem more natural to students in Utopia.

Smoothness also has a kind of cultural weight and place in Utopian discourse. Even when a natural ontology is obvious (e.g. species, genders, celestial bodies), Utopians like to push at the edge cases and explicitly flag the way that frameworks are imperfect. As a result, questions like whether “there are two genders” or “Pluto is a planet” get less traction at the object-level and instead tend to produce a common response of “yes, thank you for reminding us that all categories are imperfect.”

The Utopian emphasis on numeracy can be seen as another practical consequence of having internalized the principle of smoothness on a social level. Utopians love to measure things, and have a somewhat quirky affinity for real-valued measurements over discrete counts. For instance, it is more typical in Utopia to see groceries priced by mass, volume, or nutrients (eg calories7), rather than by count (even when they’re sold in fixed-size containers).

Prerequisites: Surveillance, Gun Control, Part 1, and Part 2. Also builds on the ideas of several other essays, such as Speed Limits.

In the first two essays in this series I think I’ve taken a somewhat middle-of-the road position on policing — appreciating the importance and historical context of police while exploring the risks and tradeoffs involved in dramatic change. I think this kind of centrism is broadly good, especially at respecting the limits of our knowledge and the complexity of the subject. And it is not incompatible with reform — our explorations have suggested several possible ways to improve police work on the margins. But in this essay I want to break a bit from conservatism and the status-quo and envision a suite of reforms that seem possible, at least in Utopia.

Let’s start by revisiting ideas from previous essays in the context of police work. For example, I believe that Utopia handles victimless crimes like drugs and prostitution not through paternalistic bans, but instead through licenses — if you want to snort cocaine, simply prove to your local community that you have your act together and understand the risks. The one major potential exception to this is gun ownership, which is totally banned in Utopia. These two changes by themselves seem like they’d already produce a huge contrast with police work in the USA. Utopian police treat it as a given that people don’t have guns and that if they have drugs, they have a license to use them.1 This, in conjunction with principled objections to searching, means Utopian police don’t engage in anything like stop and frisk, and only search people when it’s directly relevant to a known crime, such as searching a computer for records of embezzlement.

Utopian police are generally held to a high standard of honesty and transparency. Sting operations are illegal, as are undercover officers, unmarked vehicles, and other forms of secret police. Police in Utopia surveil the populace in a variety of ways, but always in an explicitly marked way, where you can clearly see you’re being watched. Outside of desk jobs, all police are equipped with a pair of body cameras that have indicator lights to show they’re active. Police are not considered to be on duty unless they’re in uniform and either in a police station or have both cameras visibly active. Off-duty police have no special powers or protections.2

Utopian surveillance extends to tracking all motor vehicles on public roads. Thanks to automation, this includes tracking vehicle speeds, allowing fines for speeding to be assigned automatically, without any police contact. Traffic police still exist in Utopia, but they’re concerned entirely with preventing reckless driving and catching those who try and evade the vehicle surveillance system. Fines associated with motor vehicles, like all legal fines, are paid out to the community, rather than directly to the police, cutting down on predatory behavior. In general, police in Utopia are funded by taxes on the unimproved value of land.

Surveillance footage and other data collected by the police is not generally accessible to officers, but can only be accessed (usually by detectives) in cooperation with a civilian oversight organization, and can only be shared when relevant for a crime and with irrelevant personal details removed. Private surveillance and security is usually illegal in Utopia, with corporations sponsoring police units to watch and guard their properties, when necessary.

Roles

Utopia doesn’t really have a single notion of “police officer” much as our world mostly doesn’t have a notion of combined policing-and-fire-fighting. Instead, most urban areas have a variety of jobs, including:

Sentry

Sentries are the most common form of Utopian police officer, tasked almost exclusively with the job of watching communities and helping those in need. Sentries patrol overlapping territories in random patterns, trying to simultaneously watch for crimes-in-progress and for people who need help of any kind, calling for backup when needed. Sentries also perform wellness checks and occasionally do acts of service for people in their community, such as helping people with physical labor or cleaning public spaces. Our world’s equivalent of security guards and traffic cops are sentries. The most common vehicle for a sentry is an electric bicycle, and sentries pride themselves on being friendly and approachable. While sentries can intervene if someone is being violent, they are generally trained to wait for backup, and are not expected to get involved in conflicts.

Peacekeeper

Peacekeepers are the most common form of dispatched Utopian police. If your neighbors are making a racket in the middle of the night, someone’s spraying graffiti, there’s a car accident, or it seems like someone might be abusing a child, Utopia sends a peacekeeper (and maybe an ambulance, if the car accident is bad). Peacekeepers are trained first and foremost as conflict mediators, focusing on empathy and nonviolent resolutions. Peacekeepers are still police, and are authorized to make arrests when necessary, but will generally avoid physical conflicts and are understood to prioritize helping community members come to agreements with each other, rather than needing to charge anyone with breaking the law.

Crisis Responder

In urban areas of Utopia, there is a distinction between peacekeepers and crisis responders. A crisis responder is an officer who is primarily dispatched to deal with someone who is in acute need, either because they are having a mental health breakdown or because they were recently the victim of a horrible crime. Where peacekeepers focus on conflict resolution between citizens, crisis responders are more like mental paramedics, handling the immediate challenge of someone who is suicidal or having a manic episode in the middle of the street. Crisis responders are trained to gather relevant information from victims in a way that helps other police roles do their jobs while also being gentle with someone who might be going through the worst night of their life.

Subduer

When a potentially violent criminal is identified and an arrest is needed, Utopia prefers to send subduers. Subduers are specialists who focus on making arrests with minimal force and maximum politeness. Unlike the above roles, however, subduers are not selected for being able to help the general populace, and are almost always large men with an interest in martial arts and an affinity for physical conflict. In small communities subduers might patrol as sentries when not actively subduing a suspect, but in larger urban areas subduers spend their downtime training, and focus entirely on the art of catching criminals without risking physical harm.

Defender

Where a subduer trains for the art of the takedown, a defender trains for physical conflicts that are not primarily about catching someone. Some defenders sub-specialize as bodyguards, site-security, blockades, or riot control. Like subduers, politeness and respect are an explicit part of the job for a defender. Unlike subduers, defenders are often free to travel to a distant place of need, rather than be on call in a particular jurisdiction. Bailiff is a job in our world that would be a defender, in Utopia.

Investigator

Investigators are nearly identical to our world’s detectives, albeit with less experience doing patrol or other police work. Investigators often train specifically for their role in school, and further specialize in various ways, such as on property crime or homicide. The best investigators often travel for work, operating across entire continents, rather than being restricted to a single city.

Other Roles

There are, of course, a variety of other roles that are connected to doing police work, such as being a dispatch operator, crime scene technician, or office clerk. Like our world, these people are not generally understood to be police, because unlike the roles explored above, they aren’t given the authority to use coercive force or make arrests.3 But unlike our world, there’s also generally an understanding that police chiefs and other leaders/administrators are not police officers. It’s still typical for a senior officer to promote into a leadership position, but in Utopia it’s also possible to become the administrator in charge of a police department while never having carried a badge, and police that become administrators are asked to turn in their badge and uniform as part of leaving their old role.

Deadly Weapons

Thanks to universal gun control, all police in utopia are also prohibited from carrying deadly weapons, including roles that, in some sense, specialize in combat. Many officers, including sentries and peacekeepers, carry pepper-spray, subduers are usually armed with electric weapons4 (e.g. tasers), and defenders sometimes wield batons, but in general there is an understanding that police officers are not soldiers and should never be armed with weapons designed to kill.

To further emphasize that police aren’t in the business of killing people, it is the norm in Utopia for any officer who kills someone (accidentally or not) to lose their job and be barred from working in any police role for over 16 years, in addition to likely being prosecuted for the homicide. Due to a variety of factors, this almost never happens in Utopia, and doesn’t have a significant impact on retaining good officers.5 Banning killers working as police is widely understood not to be a punishment (that’s up to the court), but as an assurance that when someone is interacting with a police officer, they know they’re not dealing with someone with a history of lethal force.

The removal of deadly weapons from police does not, however, mean the removal of body armor. Many police officers dress in a way that protects them in one way or another, and assists in combat, including subduers, and of course, defenders. Tower-shields are common for defenders, and for subduers who work in tactical squads. Some of this equipment can be comparable to military-grade gear. Sentries sometimes wear light armor while out on patrol in dangerous neighborhoods, but there’s an overarching pressure for most police to be mostly unarmored as part of the narrative that they are not looking to get in fights.

But what if a criminal (or gang of criminals!) gets their hands on some guns (or other crowd-killer weapons, like bombs)? Well, for starters the police are trained to command them to disarm and warn them that they are crossing a clear and bright line with regards to the use of force. If the criminal fires their gun at someone or otherwise refuses to disarm, the understanding is that they are identifying themselves not merely as a common criminal, but as a soldier. “Guns are weapons of war,” says Utopia, and if you insist on using a gun against civilization, well, civilization is allowed to call in actual soldiers to deal with you.6

Where Utopian police are trained to use the minimal amount of necessary force, Utopian soldiers are under no such obligation. After being authorized to use deadly force by the police, the army is empowered to straightforwardly kill anyone (armed or not) whom they assess is a threat, using all tools at their disposal, including heavy-duty war machines. The soldiers are, of course, instructed not to harm bystanders, but Utopia generally holds the military to a lower standard than the police.

The idea behind the use of overwhelming force in such situations is to make it clear to criminals that escalating the arms race will not be tolerated, and to prevent these situations from arising in the first place.7 A criminal armed with a knife, sword, or bow, only has to worry about getting tased or beaten, even if they’ve killed an officer with their weapon, whereas a mugger with a handgun runs the risk of getting shot by a sniper. As a result, only the most foolhardy criminals in Utopia seek out firearms, which helps eliminate the black-market for such weapons and generally solidify the taboo.

Following each instance where the police called in military support, an investigation is launched by a third-party auditor to review surveillance footage and testimony to ensure that the bright line was actually crossed. If the military is called in a context that doesn’t justify, in the eyes of the civilian auditor, the involvement of the army, the authorizing police officer is generally held responsible for the death of the target(s), meaning the Utopian police tend to err on the side of caution. (Collateral damage, including civilians caught in the crossfire, is still the army’s responsibility.)

Incentive Gradients

Utopia invests heavily in police, even though crime rates are generally lower than in our world.8 In particular, Utopia tries to prioritize police presence and responsiveness. This means having sentries visible much of the time (in urban areas, anyway) and having fast response times for peacekeepers, crisis responders, and sometimes for subduers. But it also means having investigators work around-the-clock to solve crimes and arrest criminals quickly. This is because evidence suggests that not only do cases get harder to solve as time goes on, but that increasing the probability of being caught quickly (and then sentenced) is a far stronger deterrent for would-be criminals than harshness of potential punishment. And of course, deterrence is the ideal outcome.

Everything from fast responses to polite and respectful conduct are achieved through cleverly designed organizational systems and careful use of incentives. While the full competition of the private sector is unworkable, Utopian police do still compete with each other to be the best at their roles. When possible, departments are organized to have multiple parties operating in similar conditions (same territory, etc.); these teams are then assigned randomly to cases and tasks. For instance, if there’s a break-in, a crisis response unit (usually two responders, ideally both a man and a woman) is randomly selected and dispatched to handle the immediate situation. Once it’s confirmed to be a property crime, a specialized detective is randomly selected from the pool to investigate. If the detective identifies a suspect, a random subduer unit (usually two men) is selected to make the arrest.

Later, each party can be evaluated by a third-party,9 judged only on their role in the broader system. The third-parties who are selected to evaluate police performance are also randomly selected, using a secure, internet-based service hosted by the government that picks out teams of one civilian expert and one police officer in the corresponding role to review the relevant body-cam footage and other data,10 and then score the police on relevant metrics such as how active they were, whether they used the right level of force, and so on. Response teams, detectives, etc. are then given prizes and additional funding, based on their performance, both compared to local peers and to comparable teams in other parts of the world.11

Funding for a police department as a whole (and its leadership) is usually handled by a futarchy, whether at the local level, or at the level of the world government. An example of how this works is that the futarchy sets up multiple prediction markets that pay out based on an aggregate crime statistic (e.g. violent crimes + 0.1*nonviolent crimes), where each market is conditional on the police department receiving a certain level of funding. Once the predictors weigh in, the futarchy selects the level of funding based on how the populace has voted to trade off between price and crime. This incentive means that police departments are mainly selected for keeping crime down, rather than bad proxies for efficacy like having a high number of arrests.

One final, but crucial note about law enforcement incentives is that in Utopia, when someone is found not guilty of a crime, they are automatically entitled to government compensation for all harm and inconvenience of the arrest. These costs can be significant, and serve as a major top-down pressure on police to avoid making spurious arrests. (Resisting arrest is still a crime, so if the footage indicates someone fled rather than cooperating, they can lose some or all of their compensation.) My hope is to write more about the judicial process in a future essay, and I’ll revisit this idea there.

Utopian Police

On a cultural level, Utopia regards police, especially sentries, as respected servants of the communities they protect. Police departments deliberately cultivate cultures of politeness, respect, honor, and stoicism. Utopia works hard to encourage a virtuous cycle here, where the most high-integrity men and women become police, thus holding their peers to a high standard which encourages people with an affinity for those virtues to join up. When Utopians have negative opinions of cops, it’s because they’re seen as rigid, righteous, and serious, not brutal or callous.

People in high-stress positions such as dispatch and crisis response are encouraged to either switch careers “early” or work part-time and perhaps spend their other hours on more wholesome activities like manual labor or teaching. Subduers are treated similarly, albeit with hard term limits instead of soft norms, forcing them to go into leadership, auditing, or instruction as they get older.

In general there's a lot of overlap between teachers and police roles in Utopia, especially in martial arts and spiritual mentorship. Many police see themselves as not just in charge of physically protecting their community, but also serving as role models and paragons of honor and duty who inspire right-action.

It’s normal for teenagers in Utopia, especially young men, to apply for a job as a sentry and serve with the police for a year or two. For many young people, this is their first job, and a place where they learn virtues of hard work and responsibility. Other police roles primarily recruit by offering promising young sentries an opportunity to apprentice in their departments, though of course it is also common for people to apply directly to specialized roles.

Though holidays in Utopia are varied, many communities explicitly celebrate a Police Day (sometimes combined with celebrating other first-responders like firefighters), where kids are invited to dress up as cops and criminals, play team games similar to capture the flag, and spend time in police stations interacting with the men and women who keep their homes safe.

P.S. Special thanks to Eliezer Yudkowsky’s article on Rebooting the Police for inspiration.

Prerequisites: Part 1
Foundational to: Part 3

Police (in the Western world) can trace their roots, in a certain sense, to slavery. Specifically, some time around 500 BCE, the city of Athens likely established a force of hundreds of slaves (demosioi) called “Scythian archers,”1 who were owned by the state and tasked with maintaining order and following the directives of magistrates, which could include making arrests and managing crowds during events such as public punishments. Another classical example comes from 6 CE, where we have records that Octavian established the Vigiles Urbani — an institution of watchmen, also probably composed of slaves at its founding.2 Policing is a hard job, and the use of slaves reflects the way in which many occupations that we would now describe as “blue collar” were handled by slaves in those contexts.

While the Scythian archers were likely focused on public order, the Vigiles were, most centrally, firefighters. Fire was a huge threat in ancient cities, and by patrolling the streets the watchmen had some hope of being able to spot fires early and taking action before they became catastrophic.3 To boost the ranks of the Vigiles, former slaves (libertini) were hired for pay and given citizenship after six years of service. And as the ranks of the watchmen grew, so did their duties. By 27 CE the Vigiles were responsible for spotting and arresting criminals, capturing runaway slaves, and assisting other Roman police forces4 with their duties.

The Vigiles and the Scythian archers were, in many ways, exceptional. In most historical times and places, communities were simply too small and too poor to warrant or sustain anything like a full-time police force. The able-bodied men of a community, sometimes under the direction of nobility or a local warrior, would simply need to put down their work and do their best to catch and punish wandering criminals or members of the community who broke the law. If things in a country were going particularly well, the king’s soldiers might travel the roads, hunting bandits. But in general, law and order was a question of community norms and enforcement.

Where cities managed to grow to scale, however, more examples of pre-industrial police emerged, such as the berrovarii of 1200s Bologna. Like the Scythian archers, the berrovarii were largely outsiders, drawn to Italy from places as remote as Scotland and Africa. Their outsider status helped ensure impartiality to any of the noble families of the city, who couldn’t trust each other to enforce the law in a fair way. Like modern police, the berrovarii earned a salary, carried weapons, and wore uniforms. And like the Vigiles, they held many duties, such as patrolling the streets at night to enforce curfew, confiscating illegal weapons, and breaking up gambling rings.

It’s common in classrooms and casual summaries to frame policing as a uniquely modern phenomenon — often citing Sir Robert Peel’s work establishing the Metropolitan Police Force of London in 1829 as the origin of policing as we know it.5 But even in England there are obvious precursors, such as the Bow Street Runners (1749) and Marine Police (1798). Worse are the narratives that imply that “modern” policing was somehow an American invention, perhaps being developed in 1838 Boston, or as a natural evolution of slave-catching patrols in the American South. While each place has its own history, policing itself has evolved many times and in convergent patterns, due to the underlying social needs of urban life. Whether the police are slave-catchers or slaves, whether they fight fires or enforce curfews, the core functions of watching for crime and keeping the peace have remained a consistent part of the job wherever population density and wealth are sufficiently high.

Specialization

I find the Vigiles in particular to be a somewhat fascinating case of historical policing because of how they were primarily set up to handle fires, and only ended up policing as an afterthought. The dual role makes sense, in a way — strong watchmen who are out patrolling at night seem like they’d be useful for fighting fire and for fighting crime. And yet, virtually no modern society combines things in this way. Part of that is because of how technology changed fire-fighting, but part of it is because our cities are significantly more populous and wealthy. The few parts of the world where “public safety officers” exist, jointly responsible for fire and crime, are almost always small, rural communities that can’t afford to have a dedicated fire department and police department.

Specialization is useful. Once a city grows large enough to support it, civil servants can specialize in particular domains, carrying equipment that is narrowly useful for those problems and training exclusively for their focus. Civil ambulances first started to emerge in the 1800s, but it wasn’t until the 1960s that paramedics and EMTs became a meaningful role. In smaller communities, transportation of people to a hospital is still often done by general fire-fighters and police, but if you can get handled by an advanced paramedic, that’s obviously better.6 (Interestingly, it’s hard to statistically measure how much better because of a combination of (1) response time being extremely important — getting to a hospital sooner is often much more important than who transports you — and (2) Simpson’s paradox — having a first-responder with medical training is good news, but being dispatched a medically-trained first-responder is bad news, since it means you’re clearly in medical danger.)

The USA is currently going through another wave of specialization. Driven by the anti-police brutality movements of recent years, many cities have begun experimenting with unbundling police roles that center around violent confrontation from those where an unarmed specialist might be better suited, such as mental health crises. The most famous of these programs is probably CAHOOTS (established in Eugene, Oregon, in 1989), that handles a small fraction of emergency calls by sending medics and crisis workers to handle behavioral health issues, substance use, and homelessness.

Even skeptical voices agree that programs like this are probably good, on the margin. Armed police trained in the use of violent force can be intimidating, especially to people with mental health problems, and every hour spent training to catch a thief is an hour that could have been spent learning how to talk someone down from a ledge (or vice-versa). Even when police are also present, due to a need for coercive force, specialists who have trained in empathy and conflict resolution can be invaluable, as we can see in hostage negotiations.

In most places, police handle suicide threats, homeless people, vandalism, noise complaints, hostage negotiations, public drunkenness, assault, murder, rioting, accidents, wellness checks, surveillance, traffic stops, and dozens of other responsibilities. And for the most part, police officers simply handle each of these things with some degree of competence. In small communities, there is little option but to depend on generalists, but we shouldn’t be surprised if specialized units can outperform generalists, whether they count as police or not.

Investigators and Warriors

Speaking of “counting as police,” I want to highlight a couple other interesting cases where police are, from my perspective, under-specialized, beyond the community-health and crisis workers that have been the focus of reformers in recent years.

Let’s start with detectives. In almost all cities, detectives are police officers who not only have the powers and privileges of street cops, but also started their careers doing patrols in uniform. In most places this isn’t just a typical career pathway, but an explicit requirement — detectives are required to serve on the beat for several years before being allowed to even take their entrance exam.

This is almost certainly wasted motion. While experience on patrol is undoubtedly useful, the skills and talents that make someone good at one job or the other are different. Someone might, for instance, lack the temperament or physical strength for patrol work, but possess analytical skills that make for an excellent investigator. In the UK applying directly to become a detective became an option in 2017, and federal investigators in the USA, such as the FBI, do not generally require police experience as a prerequisite. Instead of spending years on the street, serving as a junior detective assisting a more senior investigator is likely a much more efficient path.

In a similar way, consider police tactical units, widely recognized in America by the term SWAT. These specialized teams emerged in the late 1960s, driven by the need to handle high-risk situations such as hostage rescues, active shooters, or heavily armed criminals, that exceeded the capacity of regular patrol officers. Events like the Watts riots in Los Angeles underscored the necessity for such dedicated training and equipment.

Like detectives, most tactical units draw their members from regular patrol officers. But the skillsets needed for these roles diverge significantly from routine policing. SWAT operations require military-like coordination and specialized weapons training. These skills have little overlap with community policing, traffic enforcement, or investigating property crimes.

Specialization is costly, and currently, most jurisdictions with SWAT teams maintain their own tactical capabilities, even though high-risk situations requiring extreme measures are relatively rare in any given location. We might consider whether some tactical specializations could function more effectively as regional or even national resources — similar to how wildland firefighters deploy across jurisdictions. During major forest fires, specialized teams like hotshot crews travel to wherever the crisis demands, rather than each community maintaining full capabilities for rare but catastrophic events.

The militarization of police has received substantial criticism in recent years, particularly when tactical equipment and approaches are used in situations that don't warrant such force. But this doesn't mean tactical units themselves are unnecessary — rather, it suggests we need clearer boundaries around when and how these specialized capabilities should be deployed.

The Golden Mean

As we look toward improving policing (or making any change to the world, for that matter), we would be wise to keep in mind that tradeoffs exist. For all the benefits of specialization, there are a range of different benefits to generalization. Part of why police forces are so generalized is that they are the ones that have to handle uncertain situations where authority and flexibility are key to handling whatever emerges.

People having mental health crises are sometimes violent. Detectives can find themselves needing to make arrests in the course of conducting interviews. And SWAT teams from a local community may think twice before shooting first and asking questions afterward. All these considerations should be mixed with an appreciation for the limited budgets and training resources that most departments have to work for, as we push for reform.

In Part 3, I'll explore my vision of Utopia, and how a better world might handle the various duties of the police.

Prerequisites: None
Foundational to: Part 2 (and Part 3)

In the summer of 2020, set off by the murder of George Floyd by the Minneapolis police, and amplified by several other high profile deaths (particularly of African Americans) at the hands of law enforcement, massive, anti-police protests erupted worldwide. At their height, the loudest voices1 called to “defund the police” including radical abolition of the institution.

This idea should be taken seriously. The most important function of the government is to establish a monopoly on force and use that power to keep people safe, allowing them to let their guard down. If the police were simply inefficient, that would be sad but understandable — absence of competition naturally fosters sluggishness and incompetence. If these were rare cases of brutality stemming from “bad apples” that slip through the screening process, we might simply tell ourselves that the cost of “never” is (almost always) too high.

But to have repeated, horrific, high-profile examples of police officers from across the country attacking and killing civilians, including children, especially in a way that reflects deeper, systemic problems,2 should make everyone stop and reflect on whether, perhaps, this medicine is worse than the disease.

We should, at the very least, hold these victims3 in mind as an ongoing symbol of how we must do better, as a society.

Pullback

In the wake of the protests, the level of policing decreased in various ways. We can see this in metrics like the number of police stops or arrests, especially in cities that were central to the protests:

This pullback was probably due to a combination of factors,4 including decreases in funding, reduced morale leading to resignations, wanting to respond to the will of the people (including because the mayor pressures the police to lay low), and a reduction in how often the police were called to handle crimes. Any particular factor can be disputed, but the general trend is clear — there was a dramatic reduction in policing.

Sadly, it’s also not hard to see the impact of this reduced police activity. When the police pull back, crime rises sharply, especially in predominantly Black neighborhoods. The FBI’s reported national murder rate for 2020 (the year with by far the largest anti-police protests) was 30% higher than for 2019 — the largest year-on-year increase on record. Similar trends can be seen in other crime statistics, and in other times/places.

In the USA, police kill about 1250 people each year in total (including when officers use deadly force to defend themselves/bystanders5). There were about 16,700 homicides in the USA in 2019, and about 21,500 in 2020. (The rate shot up in 2020 and stayed high for about three years, but is now back to the mid 2010s level.) So if we suppose that only a quarter of that increase is due to reduced levels of policing,6 then reductions in policing in 2020 led to about the same number of deaths as all deaths at the hands of police — and far more than the deaths of innocents due to police brutality.

On a basic level, if you care about protecting people’s lives, especially those of Black people, calling for decreased levels of policing is deeply counterproductive. Police, in general, save lives, even though there are some terrible counterexamples. If anything, it seems to me that it would be wise to broadly increase police presence in areas of high crime. After all, the USA averages only ~240 police officers per 100k people — far lower than the ~340 in Western Europe, despite having a much higher rate of violent crime.

Still, just because police are important to keeping people safe does not mean we should be complacent, or simply accept occasional police brutality as unavoidable. Police behave badly for a variety of reasons. Sometimes they make mistakes that could be avoided with better training, being less overworked, or having better mental health support. Sometimes they respond violently to protect themselves in situations that they’d be more able to handle with a partner or more backup. And sometimes government can’t adequately select for high-quality officers due to upstream factors,7 bad incentives, local politics, and negotiating friction with police unions.

Market Competition

So how might we reform policing? A running theme of this blog is that society can frequently be improved by leaning more into market structures, money, and competitive self-interest. In this same spirit, we should seriously consider whether it’s possible to dramatically improve society by switching from centralized police to private companies that perform the same functions.

The primary thing that sets police apart from other groups, such as social workers or private security guards, is that police officers are given permission to use coercive force to do their jobs. But in our current society this power is essentially a monopoly. If the police aren’t doing their job, you can’t really fix that except through influencing the government. In a more ideal-market context, if you’re not satisfied with what you’re getting, you can hire a different company. This market dynamic, applied to policing, would probably lead to increased focus on efficiency, promptness, and customer satisfaction.

Some critics of this type of proposal might point out that this sounds an awful lot like how things look in parts of the world where organized crime becomes so powerful that it becomes the de facto law of the land. Are ordinary people in these places, subject to gangs and paying “protection” money, best thought of as “satisfied customers?”

But this criticism is unfair. The market-based solution is not actually an end to the notion of a monopoly on force at the high-level — we can very easily imagine a state government (or world government) that hands out legal permission to use violence to companies that abide by regulations. The government could, for example, make it illegal for private police to use their coercive privileges to, say, drive out competitors from a region and thereby extract rents from the populace.

Predatory Incentives

But even with competent government regulation, privatizing police would bring a swarm of nasty problems.

Let’s start with the most basic: creating an incentive for crime. If you have to pay a bodyguard to protect you from assault, then this means the bodyguard, on some level, benefits from the existence of would-be assailants. Even if the private police don’t actually encourage criminal behavior, either because it’s illegal to do so, or because they’re paid on a subscription/insurance model rather than in response to crime, it would still be, at the very least, in their interest to make the world seem scary.

At the start of my essay on surveillance, I discuss why this is very bad: the point of civilization is to be able to let one’s guard down and focus on vulnerable cooperation, rather than burning resources on defense. In other words, the better the government is at keeping the peace, the more it will seem like there’s no need for police. Competitive pressures, especially if advertising is allowed, would make this worse, not better.

Another nasty problem with privatizing police is distributional inequality. It’s already the case that the worst crime problems stem from the poorest neighborhoods. Market-based policing would concentrate efforts further, leaving those who can’t afford protection without recourse, while the elite attempt to shield themselves from the wave of crime afflicting the poor. Crime is, in this way, a bit like disease and policing is like vaccination. It’s most efficient, even from the perspective of the rich, to ensure laws are enforced everywhere, thus giving criminal activity nowhere to fester. (Indeed, this is part of the problem with revolving-door prisons!) A world where police resources are concentrated in the safest places is a world that is broadly unsafe, and represents a failure of civilization.

But perhaps the nastiest collection of problems from privatizing police is the way it would incentivize cops to prey on “criminals.” The sorts of police brutality we were discussing a moment ago didn’t arise because the police were attacking random people on the street — they mainly stem from interactions with suspected criminals where there was some attempt to resist or evade arrest. What better way to impress your customers than to showcase how “tough on crime” you are? What better way to cut down on costs than to train officers to follow the least-expensive path (read: the one that leads to the fewest officer injuries), rather than the path of minimal force?

These are common and pernicious problems even in our world, and they’d surely get even worse in a world with greater market pressure. Even when police forces are mostly funded through property taxes and other secular sources, all it takes is one foolish policy to lead to parasitism. Speed traps, civil forfeiture, quotas, bail bonds, and impound lots are all examples of predatory arrangements that we should be trying to minimize, not amplify.

Interlocking Systems

As we think about police reform from various angles, I think it’s important to recognize that police departments are not isolated institutions, and that many of the dynamics that characterize the government’s use of force are a result of factors that are distinct from the behavior of individual cops, or even police departments as a whole. For instance, the prevalence of guns in the USA, and the culture of violence that surrounds those guns, makes policing American streets a unique challenge.

Along those same lines, the “War on Drugs” that began in the 1970s meant high levels of top-down pressure on police to treat non-violent citizens as hardened criminals, giving rise to massive black markets which fund organized crime. Enforcing drug laws diverts resources away from violent crime and places officers in adversarial relationships with the communities they serve, particularly in poor and minority neighborhoods where drug enforcement has been most aggressive.

Regardless of whether it’s good for hard drugs to be illegal, we should recognize that improving the relationships between police and their communities is not as simple as demanding that police behave better — most of the problems are systemic, and spread into surrounding policies and institutions. No amount of careful action by law enforcement can, in isolation, heal the deep wounds of colonialism, slavery, and segregation. In these essays I’m deliberately trying to keep my focus on policing here, but true reform demands changing policy, courts, prisons, and tackling crime at its source by providing better opportunities and resources for those who are born into bad circumstances and are most at risk of turning against civilization.

Regardless, I hope you’ll join me in Part 2 and Part 3, as I discuss the historical context of police, and more promising paths for improvement towards Utopia.