Epistemic Status: My read. The mind of another is always tricky. At a minimum, Tyler would not agree with many of my framing and word choices.
The Book: The Complacent Class
Slide Deck: The Complacent Class and the Philosophy of Tyler Cowen, by my friend Anthony. This provides a good summary of his main facts and points on the surface level.
Tyler Cowen explicitly believes, and has said many times, that for society to succeed we need to believe untrue things. In his model, a society’s beliefs need to be chosen for their impact on society first. A lot of that impact is in raising the status of behaviors we want to see more, and lowering the status of behaviors we want to see less. He is often explicit about this.
Accuracy is a secondary consideration.
He might (not his example) prefer we think french fries are not delicious. We would eat less of them and be better off, even if french fries are delicious.
This is distinct from the type of self-deception Robin Hanson talks about in his book The Elephant in the Brain.
This approach is tricky. How do you avoid believing arbitrary false facts? How do we choose the actions with the best impact without an accurate model of the world? How do we engage in chains of reasoning?
This is a complex problem.
His solution is the Straussian reading.
The surface text has the first level message we want people to believe and act on, to make the world a better place. We must look deeper for the real, often opposite, second level message. We can then have a third level well-hidden discourse about what the proper first level message ought to be.
Tyler explicitly endorses looking for Straussian readings in a wide variety of texts. His philosophy implies the need for us to believe that which is not, while simultaneously being able to think objectively about the consequences of actions.
Tyler even links enthusiastically to many reviews of his book that claim his book is saying things Tyler explicitly disavows! In particular, he highlighted this truly epic masterpiece of extrapolation.
Tyler intending a Straussian reading of his own books is not only not a strange hypothesis. It should be our prior.
His first level message is that you (and America) are too complacent. You should not be complacent. To survive we must become less complacent.
His second level message is that while it is bad for society if its people are complacent, it is in your interest to be complacent.
His third level message is that we must solve this collective action problem. We must enforce and reward the norms that we need everyone to have, rather than enforce and reward those who do what causes that particular person to enjoy the best outcome.
Tyler Cowen, intentionally or not, is making the case against Causal Decision Theory.
(Important Note: Statements that I believe are labeled as such, otherwise any views expressed are my model of Tyler’s views.)
What is Complacency?
Exploitation, as opposed to exploration. Also exploitation in the sense of enjoying the benefits of our civilization without providing for its upkeep, or the upkeep of its norms, or the honoring of those who pay these upkeep costs.
The opposite of complacency is striving. A striver values exploration. Even if a striver is not getting their hands dirty on a particular issue, they honor those who do, and those who give such honor. They honor those who fight in the arena, even (especially!) when they take make mistakes or take necessary but unsavory actions.
From a selfish perspective, almost all individuals in America should mostly exploit rather than explore or strive. Exploitation has it pretty great. Exploitation captures most personal benefits of exploration at a fraction of the cost.
Society benefits when we explore, but not when we exploit. We need people to explore more and exploit less, despite this making the explorer worse off.
Tyler proposes to do this partly by making exploration easier and better, but mostly by raising the status of exploration relative to exploitation.
A book called The Exploiting Class would be misunderstood, so Tyler wisely found a less loaded word.
Complacent reads as negative but not evil. This helps us accept we are too complacent, and perhaps change.
The complacent are free riders. They want the benefits of our civilization and culture, without helping maintain it. A system with too many free riders will collapse.
Payment here does not mean money. There is plenty of money.
Reinforcing the cohesion of society, of civic life, exploring the area around you, traveling, enriching those around you: All are payment. Interacting with the physical world and the people around you is payment. Sharing ideas is payment. Creating opportunities for the unexpected is payment.
Real productivity is payment. Keeping the system running. Driving the trucks, running the trains, policing the streets, staffing the army? Payment.
Creating opportunity to do real things? Creating good jobs? Excellent payment.
Research and experimentation is even better payment. Don’t forget to publish, especially negative results.
Innovation is better still.
Honor is a key payment. Raising the status of making payments is payment! As is lowering the status of those who refuse to make payments.
As is reinforcing the beliefs and norms that create payments. We need our ethos, our founding myths and our functional solutions.
We need our functional solutions even when they involve bad things. We need to do this without endorsing those bad things.
Remember Orwell’s statement that “People sleep peaceably in their beds at night only because rough men stand ready to do violence on their behalf.” Extend this to honoring such men and actions, and other less violent ones we also prefer not to think about too carefully. We get to espouse high ideals that we could never fully uphold, because we take necessary actions that violate those ideals.
Tyler believes that fully accurate historical accounts, and fully accurate accounts of the actions our government and authorities take, are destroying our national myths and cohesion. This could be catastrophically bad. We need people to believe in liberal democratic values. We also require actions and systems that directly conflict with those values to prevent civilizational collapse.
The first best solution, for the people to understand both halves, won’t work. Most people are not capable of holding both halves in their head at once. We must find a way to do both at once anyway, or history ends up being cyclical.
The good things in life undermine our willingness to pay for their upkeep.
P. C. Hodgell said: “That which can be destroyed by the truth should be.”
Tyler Cowen could not disagree more. He sees the truth destroying things and he is terrified.
You Can’t Say That Out Loud!
Our society has made talking about some of its central issues and problems quite difficult.
A loosely defined and generally expanding class of facts, thoughts and ideas are classified as Bad Things. Many of these Bad Things were consensus views as recently as a few years ago. Many of the Bad Things are believed by many or even a majority of the people. Some are obviously true, important to people’s well being and believed by basically everyone.
If you defend even one of these Bad Things even once, for a single sentence, that makes you a horrible person. Some people will literally try to destroy your entire life for it. Even if your statement is factually accurate. Even if your true statement was not a Bad Thing at the time, but retroactively becomes a Bad Thing. Even if you do so while explicitly fighting that same Bad Thing. Or condemning Bad Thing insufficiently strongly.
There are also Good Things, for which the situation is reversed.
Tyler describes problems where people value Bad Things. In some cases he wants to point out that this is in their self-interest. In others, he wants to point out that those Bad Things are necessary for society.
Thus, he needs even more obfuscation than usual.
His book contains facts. It condemns Bad Things and praises Good Things.
If all works as planned, this constructs a model that causes the man on the street to take proper actions.
The text also contains enough information to construct a second model. This second model is deniable. It is explicit about what must be raised and lowered in status. It knows some Good Things are bad, some Bad Things good. Sometimes for the individual, sometimes for society. Most importantly, the second model explains what is going on and why, but then encourages us to preach for the first model, because the second model predicts that widespread belief in the first model will have good results.
A Trip Through the Book
When read at first level, The Complacent Class is not a great book. When read at second level, it is a much better one.
Chapter 1: The Complacent Classes
Tyler identifies three classes of complacent people.
There are those who have it made. When the exploiting is great, might as well exploit. The good life is not that expensive.
There are those who dig in. Hanging on is tough, what they have is pretty good. They feel the squeeze from society and lack the spare resources to explore. They also don’t see much upside to doing so. Better to focus on holding on to what they have.
There are those who got stuck. Things are bad, without good opportunities. Striving risks lose what little they have. Why try to escape? Effective marginal tax rates for workers trying to rise out of poverty can approach 100%. For those on disability, trying to work can mean permanent loss of benefits.
Personal marginal returns to such people striving are very low. Benefits to exploitation and complacency have gone up, benefits of exploration and striving have gone down.
Everyone exploits, then complains that no one explored.
Chapter 2: Not Moving
Moving is hardcore exploration. A lot of people are stuck in dead end places with no economic future. To allocate people properly, many must move.
Americans stopped moving. That’s really bad.
Why did they stop? Moving sucks.
Part of this is unavoidable. Moving across the street eats thousands of dollars and weeks of your life. Moving across the country eats a lot more. You need new pals and new career. You leave everyone behind. All local knowledge must be rebuilt.
Other parts are avoidable. Moving endangers vital government benefits and occupational licences. endangering your livelihood. This kept a close family member of mine in New Jersey. If you have a child, moving outside the local region is often effectively illegal.
Cities like New York City and San Francisco, where people want to move, are prohibitively expensive due to building restrictions.
We stay in our jobs longer than ever, no matter the media picture. Only 21% of employees have been in their current job less than a year. Searching for a job is even less fun than it used to be, and often means a step down.
Tyler describes an experiment. Poor families were offered the chance to move to richer neighborhoods for no extra rent, on the theory it would help children. For those who moved, future income went up by a discounted value of $99,000 per child. But half the parents turned the (free, subsidized) move down!
Chapter 3: Segregation
Tyler here is referring both to segregation by race but also by income, culture or other preferences. He notes that all are increasing, because that is what people want.
We want to live near people similar to us, go to school with them, attend religious services with them, work with them, hang out with them and be friends with them. We want them to want the same products, stores, restaurants and other such things that we do, so there will be demand for them. We value those who share our values and have things in common with us.
Rich people certainly prefer to live around and interact with other rich people. This is what rich people are mostly spending their money on. The rich person version of a thing is often better, but it’s mostly expensive to exclude the non-rich, especially for housing and schooling.
Revealed preference says many of us will spend essentially all of our money to be sorted into the group spending that amount of money on the same types of things.
He condemns this sorting with many sentences like:
Most forms of segregation ultimately corrode the basis of prosperity and innovation and eat into the trust and seed capital of society.
What are these destructive side effects? Tyler mostly does not say, attempting to pass it off as obvious. The word segregation, Bad Thing par excellence, does all the work to construct the first level.
The second level is that segregation is what everyone is spending their money on. Positional goods and signaling are crowding out other spending. The case against education is in here along with the case against zoning. Even if you’re not rich, you still spend all your money to position yourself as above those with even less money to throw into a giant pit. Saying this outright would imply that the real problem is not distributional, and undermine our democratic and egalitarian and definitely not racist values, and Tyler’s perceived support of them. So he introduces this evidence for other reasons, and allows us to figure out on our own that this sorting acts as a giant tax.
Chapter 4: Innovation
Have you heard about that whole modesty debate rationalists have been having lately? The central argument is whether or not one can benefit from trying to find new ideas and think for themselves, versus trusting in experts or consensus. The fight has mostly been at the first level, where Eliezer and others (including myself) have argued that civilization is so inadequate to solving its problems that one can personally benefit from innovation on a personal level, and can improve one’s epistemic accuracy on a broader level. This is playing in hard mode, both in the sense that thinking for one’s self is hard mode, and also that claiming that one reaps net benefits directly from doing this is also hard mode.
We fight on the first level because it is increasingly fashionable to follow Causal Decision Theory and argue that actions that help one’s self directly are smart and rational and wise, and those that do not are stupid and irrational and unwise. Most would agree that there are side benefits to the group when one engages in innovation, thinking for one’s self and questioning conventional wisdom. This is how the group gets smart in the first place. But we are forced to argue on the first level, with one hand tied behind our backs, because to argue for the second level is seen as conceding the first level, and thus losing the argument, because these hard to measure benefits to others clearly should not count.
Tyler implicitly concedes the first level, in order to argue the second one. In this context, the strategy seems reasonable, since his audience is quite different than Eliezer’s, and thus requires far more modesty.
Tyler points out that lack of innovation is very bad. It hurt productivity growth and hence economic growth, leading to stagnation. He points to monopoly power on the rise, decreasing number of Americans engaged in innovation and declining productivity growth. More than anything, he points to the lack of rising living standards.
This draws the link between innovation and prosperity, both by assumption and by math. If lack of prosperity proves lack of innovation, then prosperity depends on innovation. So by using such evidence of a lack of prosperity as proof, and by sketching this proof, he shows innovation is great and must be encouraged. Monopoly power is cited as a problem, but mostly the solution is a personal message that the reader needs to innovate.
Tyler also points out that direct statistics show we are innovating less, such as there being fewer start-ups, but does not dwell on this to avoid discouraging innovation more. What Tyler leaves unsaid is that striving to innovate is not sufficiently personally rewarded, which is why we need to encourage it by raising it in status. Tyler does not want to discourage would-be innovators by pointing out is current lack of sufficient rewards. Rather, he points out that all our futures depend on such innovators, and hopes this will inspire people to inspire others or even themselves.
I continue to think that we’ve been more innovative and productive than the statistics give us credit for. The smart phone has truly been transformational, as has the internet, as has social media, as has the rise in widely available great television, movies, music and games of all kinds. The statistics don’t properly measure these gains, and thus I am a great stagnation skeptic. However, I also view these technologies as having done a lot of damage, encouraging the rise of addictive behaviors and atomization in particular, so I do think things are rather bad. We’re very differently off with smart phones and social media, but are we better off?
Tyler says he is a happiness optimist but a revenue pessimist, so it’s not clear our perspectives here are that different. Perhaps we are mostly using different frames.
Either way, we both want more innovation, as innovation is hugely socially beneficial.
Chapter 5: Matching
Tyler believes that the gains here come primarily from matching:
I submit it is from matching, which is the supreme skill of the complacent class. We spend our money and invest our time a lot better than before because of matching. Matching is in fact the new grand project of our time, and exactly how grand still remains to be seen. Still, it is likely the largest potential source of unmeasured gains in American well-being, so let’s look at it more closely.
It’s hard to be more explicit than this: We spend our money and invest our time a lot better because of matching. That seems great!
He also notes, for those who did not notice:
“Better matching,” for all its pleasures and virtues, is also in some regards uncomfortably close to the concept of “more segregation.”
“Matching” and “segregation” are indeed two words that are very well… matched.
What does that tell us about segregation?
Tyler begins with the example of music, where less revenue is raised but for $10 a month one can search through and listen to most of the best music ever recorded, along with use of a recommendation engine. While I think those engines are terrible, the ability to freely sample potential music on demand is wonderful. We forget how awful much music used to be in a world where one had fixed albums and we could not try before we buy.
Tyler’s next example is online dating. It is hard to deny that online dating has made dating a much better experience, enabling us to find much better-matched partners with much less investment, and with much less need to play destructive games along the way.
Tyler then warns about advertisers using this matching technology. I mostly find this helpful. Advertisements for random things are obnoxious. The advertisements that are worth paying to have me see are exactly the ones I want to see. Tyler points out that this is perhaps a selfish view, because as advertisers gather too much data they will become better able to hack our preferences.
We better match students to schools, residents to hospitals (using a really cool matching algorithm) and employees to jobs.
There are even matches for dogs.
Tyler claims that those who are bad at handling information suffer from such matching. I’m not so sure. Tyler gives the example of Yelp, where many look only at the misleading star ratings rather than the informative long reviews. There’s truth to that, but even Yelp’s somewhat paid-for star ratings are a huge improvement over no information, and over time better places get more business and overall quality improves. Everyone wins.
The more worrisome case is two-way matching. When employees are better matched to jobs, that is good for productivity, but it is bad for bad employees the same way Yelp reviews are bad for bad restaurants. Before, people without much to offer as employees, or dates or friends or what not, could still sometimes find good matches due to scarcity and difficulty of search. This gave them opportunity to improve, and a subsidy from the better-off to the worse-off. By improving matching, we’ve cut those down a lot, and now those who are low-quality can only pair with others who are low-quality.
Thus, we make our society effectively more unequal and less progressive, the better matching we have, and need to compensate for that.
The second level concern is that as two-way matching improves, we force people to invest more and more resources on getting better matches. Thus, the case against education and other positional goods from chapter three is being silently screamed here. The more important matching is, the more important it is to be viewed as high-quality rather than be high-quality, the less people will invest in being actually high-quality and especially in innovation. What’s the point of doing something useful if no one knows it will be useful, and thus no one wants not only to fund you, but to have anything to do with you?
Tyler makes this explicit at the end of the chapter:
Matchers gain, strivers lose.
The nominal context there is the last important point, which is that sufficiently good two-way matching systems create increasingly efficient markets for that which is matched.
If the dating pool is not well-matched, all good signs are good signs, and one should seek someone with as many good qualities as possible. If the medical residency market is not well-matched, hospitals should seek out the best medical students on every dimension, and medical students should seek the best hospitals.
If the pools are well-matched on two sides, that changes. Now the overall quality of your match has been Zeroed Out. Your date will be ‘built on one hundred points‘ in some sense. If your date is unusually smart, you should like that only if you value smarts more than most; if you value them less, you’d want a less smart date. You’re looking for a better match with things you value more than the market, rather than a better person. You still have to watch out lest you accept a match with a worse overall score, so your cognitive burden is still there, but you despair of doing much better than you ‘deserve”.
Thus Tyler’s example of dismissing a potential date for being a Red Sox fan, even if you have no opinion of the Red Sox or baseball; this is object-level harmless, but a signal of a non-ideal fit, which is death. If it was something harmless that no one else valued, it would be fully harmless, and the date would not be ruled out.
The broader problem is that matching promotes exploitation over exploration. Inefficient matching techniques, even simple ones like having to walk around and browse to find what you want, force one to encounter and evaluate new possibilities. Having matching systems that are too efficient at finding a good short-term payoff act like dumb hill climbers. Everyone gets trapped at a local maxima, never exposed to risky but exciting possibilities they might like.
I have noticed recommendation engines getting worse at this, becoming stricter hill climbers and less engines of discovery. Amazon, which gets called out by Tyler in the context of allowing us not to leave the house, is especially seems terrible at this. Their music app’s recommendations are often as simple as “Hey, I heard you listened to Ingrid Michaelson so I got you some Ingrid Michaelson to go with your Ingrid Michaelson.” Indeed I do like Ingrid Michaelson but this is not only completely unhelpful – I figured out on my own that this was an option – it’s also the opposite of exploration.
Tyler also points out that Choices are Bad, further punishing exploration.
I have a lot more to say about matching systems, rating systems and recommendation engines, and how to build and utilize them, that would be beyond the scope of the book. I hope I get to such topics, as I find them both fascinating and useful.
Chapter 6: We Stopped Rioting and Legalized Marijuana
At this point the theme is clear. Smoking marijuana is exploitation and personal consumption. Marijuana and striving do not mix. One slows down and has a good time today, at the expense of productivity, some classes of art and creativity notwithstanding. Many report good experiences, although I do not.
Rioting and speaking out is the opposite, ideally a form of altruistic punishment. Except for letting on vent rage, rioting makes everything worse today. The threat of future rioting gives us reason to build a better tomorrow, and allows us to send messages. Taking risks to advance political causes is not in one’s self interest, and if things are pretty good for you, letting others do it also seems bad. Thus, our response of effectively de-escalating riots and shutting down protests and effective free speech. Instead of costly but meaningful actions like rioting, we now have social media likes and professionally orchestrated events, which are not remotely the same thing.
We also saw a large decrease in crime, a huge quality of life benefit. I am not sure why Tyler thinks these gains may not last, other than historical perspective or fear of the consequences of slow growth or loss of jobs.
To me this chapter was mostly about adding more examples of the relevant patterns.
Chapter 7: How a Dynamic Society Looks and Feels
Tyler here points out how different it feels to be in a dynamic society where everything is changing, like China, versus modern America and its huge investments in stability. He suggests we are outsourcing our dynamism to others, so we can take the good parts without paying the costs in upheaval and uncertainty. Again, this seems like the right move for a given person, or even in the short term for the whole society, but with bad long term consequences.
Chapter 8: Political Stagnation
Our government is terrible. It is unresponsive to the needs and opinions of its people. It can’t get things done. Everything it does do costs way too much. It has locked down many of its decisions and assigned them to automatic processes and technocrats. Gridlock rules, and getting important things done has become impossible, regardless of which important things one favors, because such things require us to strive and disrupt and everyone is more interested in short term stability and avoiding disruption.
But you knew that.
Chapter 9: The Return of Chaos
Tyler’s thesis is summed up simply as:
Ultimately peace and stability must be paid for. They must be paid for with real resources, with tax revenue, and they also require the support of people.
The rest of the chapter points to places where instability seems to be emerging, and ways in which it might further emerge. I found the evidence cited to be weak; it felt like grasping at straws and choosing the best available examples. Tyler leans on campus tensions as a sign of willingness to protest and strive, which I do not think is the underlying dynamic there. His examples of what instability might look like don’t feel that convincing, or even that unstable. Perhaps he did not want these scenarios to be what people took away from the book, or talked about. I can see that being wise.
The more convincing case is simply the quote above. We don’t need to know what form the instability will take to know that, if we can’t pay for stability, instability will come. If something can’t go on forever, it won’t.
I read The Complacent Class three month ago, but found this review hard to write for the same reasons I think Tyler wrote the book. Eliezer Yudkowsky’s Inadequate Equilibria, and the discussions that resulted, attempted to tackle related issues from a different angle. Both ask why our civilization is stuck in a variety of bad equilibria. Eliezer gives an intuitive framework for understanding how it can be in everyone’s individual interest to stay stuck. Tyler gives examples where everyone is doing what is best for them personally, to be complacent and exploit rather than to strive and explore. Tyler shows that our incentives and attitudes increasingly favor this.
We have turned this despair into norms. We say complacency and modesty are wise and rational, striving and thinking are foolish and irrational. When people stubbornly refuse to worship Moloch, or they defy Ra, we call them stupid. Or worse.
How do we get to a better equilibrium?
Eliezer speaks to the individual. He says, behold the failure modes and foolishness that trap us. Even if you cannot coordinate to get us out, if you think for yourself on the object-level, you can find ways to know things, and at least improve your own life through direct action. We have to face Moloch alone, but we need not listen to his whispers in our head. The greater scale benefits of such actions are left implicit.
Tyler speaks to the society. He says, behold the failure modes and foolishness that trap us. We must coordinate to get out, and return to thinking for ourselves and fostering exploration, experimentation and interaction. Facing down Moloch is bad for the individual living among his army, but together we can best him. The first step is to stop whispering his praises, and start cursing his name. We raise the status of those who strive, and lower that of the complacent. We make people believe that striving is the path to success. He hopes some read that striving is higher status and more promising, while others read that we must raise its status and make it appear more promising. Then the meta-norm takes over, incentives change and the equilibrium shifts.
Both answers hold promise.
Eliezer is right that more exploration and doing more object-level work is good for the individual on the margin. We are falling into addictive traps and engaging in hyperbolic discounting, resulting in habits reinforcing exploitation we don’t even enjoy over exploration that would prove valuable. Eliezer’s audience has even more to gain here than most.
This alone won’t get it done, though, because Tyler is right too. Even before social pressures and hyperbolic discounts make things worse, individuals benefit from doing more exploiting than is socially optimal. Getting people to do the personally optimal mix, while ignoring the bigger picture, isn’t good enough. We need Tyler’s approach as well. Tyler admits that complacency is the self-interested thing to do, and suggests we work to fix that through shifting incentives and norms, and that we consider the bigger picture. Doing what directly personally benefits us, without an eye to what causes and is caused by such behaviors and habits and norms, is neither wise nor praiseworthy.
Is Tyler right that in order to do this, that as a society we must believe untrue things? Or could we accomplish this with a better set of norms that understood and reinforced the value of striving and the links between different people’s values and actions sufficiently to result in the right amount of striving? Can the American on the street learn to approximate functional decision theory?
I’m not sure. I suspect that we can do it. I hope that we can do it, that if those who can understand and who led us into this hole could agree to stop digging, that the resulting norms would do the rest. People want to do big things, take chances, help others. We intuitively grasp the subtle causal links across time, the honor we should bestow upon the striver, the call to adventure. I hope it is only because we have lied to them that these desires are irrational and bad, and overwritten those intuitions, that we then feel the need to lie to them again about what will directly improve their lives.