A lot of comments on my last blog post were along the lines of “but humans are political animals, we have to have some sort of ideas about how society works, however poorly grounded in evidence.” So I think I’m going to outline my thoughts on how to reason about society as a whole without use of poorly grounded sociology. First, I’m going to talk about strategies for avoiding sociology entirely when possible; second, I’m going to talk about when we have to make generalizations about society and how to do it right.

Sociology is really complicated and hard. Therefore, whenever possible, we should strive to avoid doing sociology– the study of societies– and instead try to do something else.

The amount of amateur sociology in the rationalist community is dwarfed by the amount of amateur psychology. However, the amateur psychology is much less of an epistemic madhouse.

Nearly everyone can talk, in depth, about the psychology of twenty people; very few people can talk in depth about the sociology of twenty societies. For this reason, amateur psychology is likely to have a much better evidence base to begin with. A lot of amateur sociology is in the position of an alien psychologist who wants to try to learn how humans work, but only studies the one person they abducted, Joe, without ever referencing another human. Not only is this not a great way to understand humans in general, it’s not even a great way to understand Joe.

Furthermore, unlike sociology, amateur psychology can often be tested. If someone makes a claim like “we can polyhack!” or “to avoid getting in troubles because of what words you’re using, play Taboo” or “lots of cis people don’t really have gender identities”, then other people can try polyhacking, playing taboo, or introspecting about their gender identities. If an idea turns out to not work so well (remember polyphasic sleep?) then it eventually gets dropped; if an idea works really well (Beeminder), it is more widely adopted.

It is often possible to convert your amateur sociology to something narrower. Toxoplasma of rage may or may not apply to conversations in general, but it definitely applies to conversations on Tumblr. Insightful or interesting Tumblr posts usually get fewer notes than those which prompt a flame war; political groups have a depressing tendency to be represented, in the mind of the general public, with people who are trolls and/or literally twelve. The average Tumblr user has viewed hundreds if not thousands of Tumblr conversations, and has more than enough of an evidence base to conclusively state that this is true.

However, this description is strictly about the what, rather than the why. It is agnostic about what causes Tumblr to be this way. Perhaps it is a single instance of a general tendency for human conversations to favor the controversial; perhaps it is because Tumblr’s userbase tends to be highly combative; perhaps Tumblr’s design leads to controversy, because no one can argue with a post without showing it to all of their followers, who will want to reblog a sufficiently inflammatory post to argue with themselves. We do not have enough evidence to favor one of these hypotheses over another.

Nevertheless, it is very useful. You’re not engaging in conversations in general; you’re engaging in conversations on Tumblr. You don’t have to know whether controversial things gaining population is a general trait of conversations in order to notice the trend and take steps to avoid it, such as choosing to check whether people are trolls and/or literally twelve before you reblog that takedown that TOTALLY DESTROYED them. A surprising amount of work can be done strictly on the object level.

But sometimes you can’t make the idea narrow. If in your experience a lot of your relationships with men involve them relying on you for emotional support and relationship maintenance, then there isn’t a specific narrow context you’re talking about, in the same way that toxoplasma of rage can apply to Tumblr. However, there is a single narrow context that you can apply to almost anything: your life.

If you are like “this tends to be true in my life”, you can get almost all the benefits of applying the pop-feminist concept of emotional labor (which ought to be distinguished from the actual concept of emotional labor) to your relationships. You can say: “I wonder why I feel so drained every time I interact with that guy? Oh yeah! I’m performing emotional labor as defined by pop feminists! Recognizing this pattern will allow me to set boundaries much more quickly than if I had to come up with the idea from scratch every time!”

On the other hand, once again, you’re applying the concept descriptively to a relatively narrow domain. You’re not able to show from this that men as a class feel entitled to extract labor from women in their relationships. Of course, that might be the case. But it might be that men are less likely to have emotionally intimate relationships with other men, leading them to rely on women for all their emotional needs and to have less practice in returning the favor. It might be that you have internalized sexism which makes it harder for you to set boundaries with men than with women. It might be that you don’t notice all your emotionally draining relationships with women, because you are an evil man-hating feminist. It might be that you just met a bunch of asshole guys. And you certainly can’t generalize beyond men you tend to interact with: the pattern is likely to be very different for a person from a different race, class, ability, region and/or subculture.

Okay, so, none of those things work. You are pretty sure you have to know something about actual society. My first question is: do you? Really?

Like, honestly, I don’t know anything about foreign policy and I’m not really interested in knowing much about foreign policy; I am also ignorant of health care economics, macroeconomic stabilization policy, and a dozen other topics relevant to how society works. None of this astonishing ignorance seems to have had any effect of my life, beyond annoying my friend Keller. (Sorry Keller.)

A lot of people feel bad about saying “I don’t know and I don’t care to know”. It makes them feel stupid, or uninformed, or like a bad caricature of $PoliticalParty from a political cartoon. But honestly you can’t know everything. There is no shame in admitting that you are not capable of being omniscient.

Of course, it would probably be desirable for me to actually understand foreign policy and macroeconomic stabilization and so on. But the question isn’t just “would it be beneficial for me to know something about this?”; the question is “would it be beneficial for me to replace my honest ignorance with an opinion founded on little to no evidence?” In most cases (term papers and freshman bull sessions aside), the answer is ‘no’. At least in the former case you’re keeping track of what you’re ignorant about; if it suddenly becomes important that you know about the issue, you know that you need to learn about it, instead of having cached thoughts in the back of your head.



Eliezer once called emergence the junk food of curiosity; I think insight porn also qualifies. When you read Meditations on Moloch, it feels like you’ve learned something; you get the sweet, concentrated superstimulus of understanding a dozen different topics on a new, fundamental level. But you don’t really understand them any better than you used to; your actual knowledge of corporate welfare or the rise of agriculture, your ability to predict what will happen given certain situations, is probably about where it was before.

Just like Skittles are part of a balanced food diet, insight porn is part of a balanced intellectual diet. But it should be an occasional treat! Mostly, you should try to understand topics on a deep, nuanced level: understanding the various positions and the evidence behind them, knowing what has been established firmly and what is on shaky ground, being familiar with the research. Insight porn can never replace the virtue of scholarship.

In areas where you want to have an opinion but you are not willing to research enough to have a proper opinion, I would recommend attempting to discover the expert consensus and believe it. We do this already on a lot of topics: I suspect most people who believe in evolution, global warming, or fermions could not explain to you why experts believe in those things. Conveniently, philosophers have made their expert consensus clear; as far as I’m aware, other academic fields do not have polls about their opinions about the important issues, and one instead has to figure out what the consensus opinion is by (for instance) reading introductory textbooks.

It’s important also to note when there isn’t broad consensus in a field, because this is particularly common in the social sciences and humanities. Whether it’s psychology’s replication crisis, the wide-ranging disagreement in the Philpapers survey, or the fact that theologians still have not settled which religion is right, there are many cases where the expert consensus is “fucked if I know”.

How to deal with that is where I’m getting into the tenuous “Ozy is a hypocrite who uses amateur sociology” area. There are heuristics one can use to reason about society even under conditions of very great uncertainty. One is to attempt to make big changes slowly in a reversible fashion: don’t support open borders when you can support a gradual increase in immigration; don’t support eradicating zoning when you can support lightening it; don’t support revolution when you can support reform. In this way, you are more likely to notice unexpected negative consequences before they blow everything up.

Another heuristic is to attend to the easily observable first-order consequences and leave the second-order consequences to themselves. Allowing gay people to get married provides the benefits of marriage to gay people; not misgendering trans people helps prevent dysphoria. Is it possible that LGBT acceptance will lead to the destruction of the family? Certainly! But that is highly speculative, and so should be considered less important compared to the relatively certain benefits.