The Treacherous Path to Rationality

by Jacobian10 min read9th Oct 202099 comments


Pitfalls of RationalityGroup RationalityEpistemologyRationalityCommunity

Cross-posted, as always, from Putanumonit.

Rats v. Plague

The Rationality community was never particularly focused on medicine or epidemiology. And yet, we basically got everything about COVID-19 right and did so months ahead of the majority of government officials, journalists, and supposed experts.

We started discussing the virus and raising the alarm in private back in January. By late February, as American health officials were almost unanimously downplaying the threat, we wrote posts on taking the disease seriously, buying masks, and preparing for quarantine.

Throughout March, the CDC was telling people not to wear masks and not to get tested unless displaying symptoms. At the same time, Rationalists were already covering every relevant angle, from asymptomatic transmission to the effect of viral load, to the credibility of the CDC itself. As despair and confusion reigned everywhere into the summer, Rationalists built online dashboards modeling nationwide responses and personal activity risk to let both governments and individuals make informed decisions.

This remarkable success did not go unnoticed. Before he threatened to doxx Scott Alexander and triggered a shitstorm, New York Times reporter Cade Metz interviewed me and other Rationalists mostly about how we were ahead of the curve on COVID and what others can learn from us. I told him that Rationality has a simple message: “people can use explicit reason to figure things out, but they rarely do”

If rationalists led the way in covering COVID-19, Vox brought up the rear

Rationalists have been working to promote the application of explicit reason, to “raise the sanity waterline” as it were, but with limited success. I wrote recently about success stories of rationalist improvement but I don’t think it inspired a rush to LessWrong. This post is in a way a response to my previous one. It’s about the obstacles preventing people from training and succeeding in the use of explicit reason, impediments I faced myself and saw others stumble over or turn back from. This post is a lot less sanguine about the sanity waterline’s prospects.

The Path

I recently chatted with Spencer Greenberg about teaching rationality. Spencer regularly publishes articles like 7 questions for deciding whether to trust your gut or 3 types of binary thinking you fall for. Reading him, you’d think that the main obstacle to pure reason ruling the land is lack of intellectual listicles on ways to overcome bias.

But we’ve been developing written and in-person curricula for improving your ability to reason for more than a decade. Spencer’s work is contributing to those curricula, an important task. And yet, I don’t think that people’s main failure point is in procuring educational material.

I think that people don’t want to use explicit reason. And if they want to, they fail. And if they start succeeding, they’re punished. And if they push on, they get scared. And if they gather their courage, they hurt themselves. And if they make it to the other side, their lives enriched and empowered by reason, they will forget the hard path they walked and will wonder incredulously why everyone else doesn’t try using reason for themselves.

This post is about that hard path.

The map is not the territory


Alternatives to Reason

What do I mean by explicit reason? I don’t refer merely to “System 2”, the brain’s slow, sequential, analytical, fully conscious, and effortful mode of cognition. I refer to the informed application of this type of thinking. Gathering data with real effort to find out, crunching the numbers with a grasp of the math, modeling the world with testable predictions, reflection on your thinking with an awareness of biases. Reason requires good inputs and a lot of effort.

The two main alternatives to explicit reason are intuition and social cognition.

Intuition, sometimes referred to as “System 1”, is the way your brain produces fast and automatic answers that you can’t explain. It’s how you catch a ball in flight, or get a person’s “vibe”. It’s how you tell at a glance the average length of the lines in the picture below but not the sum of their lengths. It’s what makes you fall for the laundry list of heuristics and biases that were the focus of LessWrong Rationality in the early days. Our intuition is shaped mostly by evolution and early childhood experiences.

Social cognition is the set of ideas, beliefs, and behaviors we employ to fit into, gain status in, or signal to groups of people. It’s often intuitive, but it also makes you ignore your intuition about line lengths and follow the crowd in conformity experiments. It’s often unconscious — the memes a person believes (or believes that they believe) for political expediency often just seem unquestionably true from the inside, even as they change and flow with the tides of group opinion.

Social cognition has been the main focus of Rationality in recent years, especially since the publication of The Elephant in the Brain. Social cognition is shaped by the people around you, the media you consume (especially when consumed with other people), the prevailing norms.

Rationalists got COVID right by using explicit reason. We thought probabilistically, and so took the pandemic seriously when it was merely possible, not yet certain. We did the math on exponential growth. We read research papers ourselves, trusting that science is a matter of legible knowledge and not the secret language of elevated experts in lab coats. We noticed that what is fashionable to say about COVID doesn’t track well with what is useful to model and predict COVID.

On February 28th, famous nudger Cass Sunstein told everyone that the reason they’re “more scared about COVID than they have any reason to be” is the cognitive bias of probability neglect. He talked at length about university experiments with electric shocks and gambles, but neglected to calculate any actual probabilities regarding COVID.

While Sunstein was talking about the failures of intuition, he failed entirely due to social cognition. When the article was written, prepping for COVID was associated with low-status China-hating reactionaries. The social role of progressive academics writing in progressive media was to mock them, and the good professor obliged. In February people like Sunstein mocked people for worrying about COVID in general, in March they mocked them for buying masks, in April they mocked them for hydroxychloroquine, in May for going to the beach, in June for not wearing masks. When someone’s view of COVID is shaped mostly by how their tribe mocks the outgroup, that’s social cognition.

Underperformance Swamp

The reason that intuition and social cognition are so commonly relied on is that they often work. Doing simply what feels right is usually good enough in every domain you either trained for (like playing basketball) or evolved for (like recoiling from snakes). Doing what is normal and fashionable among your peers is good enough in every domain your culture has mastered over time (like cooking techniques). It’s certainly good for your own social standing, which is often the main thing you care about.

Explicit rationality outperformed both on COVID because responding to a pandemic in the information age is a very unusual case. It’s novel and complex, long on available data and short on trustworthy analysis, abutting on many spheres of life without being adequately addressed by any one of them. In most other areas reason does not have such an inherent advantage.

Many Rationalists have a background in one of the few other domains where explicit reason outperforms, such as engineering or the exact sciences. This gives them some training in its application, training that most people lack. Schools keep talking about imparting “critical thinking skills” to all students but can scarcely point to much success. One wonders if they’re really motivated to try — will a teacher really have an easier time with 30 individual critical thinkers rather than a class of password-memorizers?

Then there’s the fact that most people engaged enough to answer a LessWrong survey score in the top percentile on IQ tests and the SAT. Quibble as you may with those tests, insofar as they measure anything at all they measure the ability to solve problems using explicit reason. And that ability varies very widely among people.

And so most people who are newly inspired to solve their problems with explicit reason fail. Doubly so since most problems people are motivated to solve are complicated and intractable to System 2 alone: making friends, losing weight, building careers, improving mental health, getting laid. And so the first step on the path to rationality is dealing with rationality’s initial failure to outperform the alternatives.

Watch out for the alligators of rationalization camouflaged as logs of rationality

Sinkholes of Sneer

Whether someone gives up after their initial failure or perseveres to try again depends on many factors: their personality, context, social encouragement or discouragement. And society tends to be discouraging of people trying to reason things out for themselves.

As Zvi wrote, applying reason to a problem, even a simple thing such as doing more of what is already working, is an implicit accusation against everyone who didn’t try it. The mere attempt implies that you think those around you were too dumb to see a solution that required no gifts or revelations from higher authority, but mere thought.

The loudest sneers of discouragement come from those who tried reason for themselves, and failed, and gave up, and declared publicly that “reason” is a futile pursuit. Anyone who succeeds where they failed indicts not merely their intelligence but their courage.

Many years ago, Eliezer wrote about trying the Shangri-La diet, a strange method based on a novel theory of metabolic “set points” and flavor-calorie dissociation. Many previous casualties of fad diets scoffed at this attempt not because they spotted a clear flaw in the Shangri-La theory, but at Eliezer’s mere hubris at trying to outsmart dieting and lose weight without applying willpower.

Oh, you think you’re so much smarter? Well let me tell you…

A person who is just starting (and mostly failing) to apply explicit reason doesn’t have confidence in their ability, and is very vulnerable to social pressure. The are likely to persevere only in a “safe space” where attempting rationality is strongly endorsed and everything else is devalued. In most normal communities the social pressure against it is simply too strong.

This is I think is the main purpose of LessWrong and the Rationalist community, and similar clubs throughout history and around the world. To outsiders it looks like a bunch of aspie nerds who severely undervalue tact, tradition, intuition, and politeness, building an awkward and exclusionary “ask culture“. They’re not entirely wrong. These norms are too skewed in favor of explicit reason to be ideal, and mature rationalists eventually shift to more “normie” norms with their friends. But the nerd norms are just skewed enough to push the aspiring rationalist to practice the craft of explicit reason, like a martial arts dojo.

Strange Status and Scary Memes

But not all is smooth sailing in the dojo, and the young rationalist must navigate strange status hierarchies and bewildering memeplexes. I’ve seen many people bounce off the Rationalist community over those two things.

On the status front, the rightful caliph of rationalists is Eliezer Yudkowsky, widely perceived outside the community to be brash, arrogant, and lacking charisma. Despite the fact of his caliphdom, arguing publicly with Eliezer is one of highest-status things a rationalist can do, while merely citing him as an authority is disrespected.

People like Scott Alexander or Gwern Branwen are likewise admired despite many people not even knowing what they look like. Attributes that form the basis of many status hierarchies are heavily discounted: wealth, social grace, credentials, beauty, number of personal friends, physical shape, humor, adherence to a particular ideology. Instead, respect often flows from disreputable hobbies such as blogging.

I think that people often don’t realize that their discomfort with rationalists comes down to this. Every person cares deeply and instinctively about respect and their standing in a community. They are distressed by status hierarchies they don’t know how to navigate.

And I’m not even mentioning the strange sexual dynamics

And if that wasn’t enough, rationalists believe some really strange things. The sentence “AI may kill all humans in the next decade, but we could live forever if we outsmart it — or freeze our brains” is enough to send most people packing.

But even less outlandish ideas cause trouble. The creator of rationality’s most famous infohazard observed that any idea can be an infohazard to someone who derives utility or status from lying about it. Any idea can be hazardous to to someone who lacks a solid epistemology to integrate it with.

In June a young woman filled out my hangout form, curious to learn more about rationality. She’s bright, scrupulously honest, and takes ideas very seriously, motivated to figure out how the world really works so that she can make it better. We spent hours and hours discussing every topic under the sun. I really liked her, and saw much to admire.

And then, three months later, she told me that she doesn’t want to spend time with me or any rationalists anymore because she picked up from us beliefs that cause her serious distress and anxiety.

This made me very sad also perplexed, since the specific ideas she mentioned seem quite benign to me. One is that IQ is real, in the sense that people differ in cognitive potential in a way that is hard to change as adults and that affects their potential to succeed in certain fields.

Another is that most discourse in politics and the culture war can be better understood as signaling, a way for people to gain acceptance and status in various tribes, than as behavior directly driven by an ideology. Hypocrisy is not an unusually damning charge, but the human default.

To me, these beliefs are entirely compatible with a normal life, a normal job, a wife, two guinea pigs, and many non-rationalist friends. At most, they make me stay away from pursuing cutting-edge academic mathematics (since I’m not smart enough) and from engaging political flame wars on Facebook (since I’m smart enough). Most rationalist believe these to some extent, and we don’t find it particularly remarkable.

But my friend found these ideas destabilizing to her self-esteem, her conception of her friends and communities, even her basic values. It’s as if they knocked out the ideological scaffolding of her personal life and replaced it with something strange and unreliable and ominous. I worried that my friend shot right past the long path of rationality and into the valley of disintegration.

Valley of Disintegration

It has been observed that some young people appear to get worse at living and at thinking straight soon after learning about rationality, biases, etc. We call it the valley of bad rationality.

I think that the root cause of this downturn is people losing touch entirely with their intuition and social cognition, replaced by trying to make or justify every single decision with explicit reasoning. This may come from being overconfident in one’s reasoning ability after a few early successes, or by anger at all the unreasoned dogma and superstition one has to unlearn.

A common symptom of the valley are bucket errors, when beliefs that don’t necessarily imply one another are entangled together. Bucket errors can cause extreme distress or make you flinch away from entire topics to protect yourself. I think this may have happened to my young friend.

My friend valued her job, and her politically progressive friends, and people in general, and making the world a better place. These may have become entangled, for example by thinking that she values her friends because their political activism is rapidly improving the world, or that she cares about people in general because they each have the potential to save the planet if they worked hard. Coming face to face with the ideas of innate ability and politics-as-signaling while holding on to these bucket errors could have resulted in a sense that her job is useless, that most people are useless, and that her friends are evil. Since those things are unthinkable, she flinched away.

Of course, one can find good explicit reasons to work hard at your job, socialize with your friends, and value each human as an individual, reasons that have little to do with grand scale world-improvement. But while this is useful to think about, it often just ends up pushing bucket errors into other dark corners of your epistemology.

People just like their friends. It simply feels right. It’s what everyone does. The way out of the valley is to not to reject this impulse for lack of journal citations but to integrate your deep and sophisticated friend-liking mental machinery with your explicit rationality and everything else.

Don’t lose your head in the valley

The way to progress in rationality is not to use explicit reason to brute-force every problem but to use it to integrate all of your mental faculties: intuition, social cognition, language sense, embodied cognition, trusted authorities, visual processing… The place to start is with the ways of thinking that served you well before you stumbled onto a rationalist blog or some other gateway into a method and community of explicit reasoners.

This idea commonly goes by metarationality, although it’s certainly present in the original Sequences as well. It’s a good description for what the Center for Applied Rationality teaches — here’s an excellent post by one of CFAR’s founders about the valley and the (meta)rational way out.

Metarationality is a topic for more than two paragraphs, perhaps for an entire lifetime. I have risen out of the valley — my life is demonstrably better than before I discovered LessWrong — and the metarationalist climb is the path I see ahead of me.

And behind me, I see all of this.

So what to make of this tortuous path? If you’re reading this you are quite likely already on it, trying to figure out how to figure things out and dealing with the obstacles and frustrations. If you’re set on the goal that this post may offer some advice to help you on your way: try again after the early failures, ignore the sneers, find a community with good norms, and don’t let the memes scare you — it all adds up to normalcy in the end. Let reason be the instrument that sharpens your other instruments, not the only tool in your arsenal.

But the difficulty of the way is mostly one of motivation, not lack of instruction. Someone not inspired to rationality won’t become so by reading about the discouragement along the way.

And that’s OK.

People’s distaste for explicit reason is not a modern invention, and yet our species is doing OK and getting along. If the average person uses explicit reason only 1% of the time, the metarationalist learns that she may up that number to 3% or 5%, not 90%. Rationality doesn’t make one a member of a different species, or superior at all tasks.

The rationalists pwned COVID, and this may certainly inspire a few people to join the tribe. As for everyone else, it’s fine if this success merely raises our public stature a tiny bit, lets people see that weirdos obsessed with explicit reason have something to contribute. Hopefully it will make folk slightly more likely to listen to the next nerd trying to tell them something using words like “likelihood ratio” and “countersignaling”.

Because if you think that COVID was really scary and our society dealt with it really poorly — boy, have we got some more things to tell you.