A coordination problem is a situation where, in order to get the best outcome, everyone needs to be doing the same thing. For example, which side of the road you drive on is a coordination problem: if everyone drives on the same side of the road, it’s okay, but if some people drive on the left and some people drive on the right, there will be car crashes everywhere. Global climate change is another example of a coordination problem. You can’t fix global climate change by deciding that you yourself will stop polluting; everyone has to cut back on their carbon emissions to prevent climate change. Lots of the problems in the world are coordination problems.

Unfortunately, some people I know claim that things are coordination problems when really they aren’t. Often, they don’t bother to try to establish that the thing is a coordination problem first; they just skip ahead to the part where they morally exhort people about the importance of not participating in coordination problems and/or complain about how the rationalist community is supposed to be full of people who are rational and yet here we are all irrationally having coordination problems.

I’m going to take social network use as my example, but I think the error occurs more generally.

In many ways, Facebook is bad. It invades your privacy. Hundreds of very smart people make very high salaries to try to figure out how to get you to keep clicking on Facebook, even when it doesn’t make you happy and you’d rather be doing literally anything else. Its algorithms often favor clickbait-y news over thoughtful longreads that give you a deep understanding of a particular issue.

If Facebook is bad, some people might wonder, why do people still use it? The “coordination problem” answer is that everyone else is on Facebook. A social network isn’t very much fun if no one is on it, so everyone has to use Facebook because everyone else is on Facebook, even though it is a bad website.

I don’t think this is true. I have personally observed a very successful shift away from a social networking website in 2007. In the Strikethrough and Boldthrough events, LiveJournal deleted a number of journals for expressing interest in sex-related topics such as BDSM, sex work, pedophilia, and rape. Infamously, this included communities that supported rape survivors; infamously, it also included a bunch of fanfiction porn communities.

In the wake of Strikethrough and Boldthrough, Archive of Our Own was founded as a free-speech alternative to LiveJournal and other ways that people commonly hosted fanfic. Today, it is one of the world’s most popular fanfiction archives, and the only known case of a free-speech alternative to something that didn’t become immediately overrun with Nazis.

By the “coordination problem” argument, there should have been a lot of trouble in getting everyone to use Archive of Our Own instead of LiveJournal. But there really wasn’t. Fanfiction writers genuinely care about Harry Potter porn free speech, and so when a site proved itself unreliable at hosting Harry Potter porn free speech, they quickly switched to a different website.

(Tumblr is unfortunately doing a similar attack on Harry Potter porn free speech, but moving seems to be going poorly due to the lack of an adequate substitute for Tumblr. I think you can get people to switch to a good substitute, but of course if the grossly inadequate website is the best you can get then people aren’t going to move.)

That places a lower bound on how hard it is to get people to move. In order to get people en masse to change from one social networking site to another, they have to care about the problems with the first site at least as much as fanfiction writers care about Harry Potter porn free speech. That’s it.

For any social network, there are three kinds of users. There are people who prefer the social network in question to any other social networks available. There are people who are kind of meh about it– this social network is fine, but there are others that are just as good. And there are people who hate the social network but can’t move because all their friends are there.

The “coordination problem” argument requires that most people are in the second or third groups, with only a small number of people in the first group. And I don’t think that’s actually true. If, say, forty percent of your users are “this social network sucks” and sixty percent are “meh, whatever, I’ll go where everyone else is going,” the forty percent can post about it, create common knowledge of the fact that they think it sucks, move to a different social networking site, and drag the apathetic sixty percent along with them.

The situation with Facebook is that there are lots of people in the first group– people who actively like Facebook. They like getting in political arguments and getting tagged in selfies and looking at pictures of other people’s babies. They don’t really care about privacy, don’t necessarily mind spending an evening on Facebook when they meant to be doing something else, and enjoy reading clickbait about which Game of Thrones character looks most like your cat. These people would continue to use Facebook even if it halved in size, as long as the Game of Thrones quizzes kept coming. In fact, I’m pretty sure the “uses Facebook, actively likes Facebook” group outnumbers the “uses Facebook, hates Facebook, wants to switch to something else” group.

Now, you can argue that those people’s preferences are stupid preferences. Probably they should care more about their privacy and being hacked into compulsive behavior they don’t endorse and should prefer reading Dostoyevsky to taking quizzes about cats. But that’s not a coordination problem. That’s a “the things people care about are bad and they should care about different things” problem. You do not fix that problem through moral exhortation to avoid coordination problems or informing people that the entire point of the rationalist community is avoiding coordination problems. And once you have fixed people’s preferences, I think the coordination problem will actually take care of itself.