A Cascade Of Dunbar Numbers

Back when I used to play at micronations, we had grand dreams of creating a fully-fleshed out simulated country with hundreds of citizens and dozens of Byzantine government departments and industries. For a while, these dreams seemed attainable. Most of the people we advertised to were really interested, and we went from one or two founding members to about a dozen people very quickly.

Then we stopped. For the past fifteen years or so, the population of Shireroth remained fixed at between ten to twenty people. It wasn’t always the same ten or twenty – some new people entered, some old people left – but it always stayed within that window.

Ten years ago we got really tired of this and spent a decent amount of money on an advertising campaign. It was a good campaign, and we got maybe thirty or forty people to apply for citizenship at once. We thought this was going to be so awesome – the size of the simulation quintupling within a week. Instead, the new people just kind of wandered around aimlessly, bumping into things, never quite figuring out what was going on, never quite taking an active role in anything or showing any initiative, finally showing up less and less until two months later the population was back down to about a dozen people.

Our Ministry of Immigration and Naturalization, which was in charge of recruiting new people, kept beating its head up against the same sort of idea with about the same results. I thought this was dumb, so I manuevered my way into the Ministry and shifted all of our resources from recruitment to retention – setting up mentoring programs for new citizens, writing a bunch of guides for what they were supposed to do, trying to befriend them and integrate them into our social scene. The result was that our new immigrants maybe stayed three months instead of two months before evaporating.

We were only ever able to find one successful method of recruiting new micronationalists: wait for random people around the world to learn about micronations, start their own projects without any impact or influence from us, integrate those projects into the larger community, and then if those projects failed after a year or two sometimes their members would join us instead. This sort of worked, but never enough to get us more than ten or twenty people. And those other projects, the ones they made, very rarely had more than ten or twenty people either, despite hordes of people who said they thought micronations were interesting and agreed to participate for a few months. It was a weird problem, and one I was never able to solve before I left the hobby.

I was thinking about this recently because of some people’s complaints about the Bay Area rationalist community.

I have always had universally positive experiences with the Bay Area rationalist community. I arrived in about 2012 when it was still smallish, and I got to know the group involved and hang out with fascinating people like Eliezer Yudkowsky and Anna Salamon and Michael Vassar and learn a lot from them.

The complaints I heard were something like this. A lot of people join rationalist communities in Missouri or Italy or Australia or some far-flung place like that and really like them and feel like they have a great tight-knit community. They decide that if a small rationalist community is good, a ten-times-larger rationalist community will be ten times as good. Then they get here and everyone’s already doing their own thing, and it turns out Eliezer and Anna and Michael do not have enough time to personally interact with every single person in San Francisco on an intimate basis, and they kind of hang around the edges of the community not really knowing what to do or being connected to its general rhythm. It doesn’t help that a lot of people (DESPITE MY BEST EFFORTS) keep thinking that rationality will turn them into supermen and get upset that they’ve been in the Bay Area like an entire month and are not yet Elon Musk.

And I was thinking of both of these things when I met Samo in Oakland and he told me about his research. It’s kind of complicated, but when I asked him “Is this basically hypothesizing a cascade of Dunbar numbers?” he told me I wasn’t far off.

Dunbar’s number, remember, is the theory of anthropoligst Robin Dunbar that humans have a hard-coded optimal group size. He found that most primates’ brain size correlates with the average size of their social groups. If you extrapolate the connection to humans, you would expect humans to have a social group of about a hundred fifty people (other estimates find anywhere between 100 and 250). Various researchers and amateurs have claimed to corroborate this in every context from hunter-gatherer bands to corporations to MMORPG guilds (1, 2), sometimes more believably than others.

Samo said his own research had found several of these discontinuities – I think he mentioned about 12 people, about 150 people, about 1,000 people, and about 90,000 people as the first few. He tried to tie this in to government forms like “family”, “clan”, “tribe”, and “city-state”. I think the idea was supposed to be that we’re naturally very good at dealing with groups of family size, our brains will grudgingly allow us to deal with groups of up to clan size, with the low-hanging fruit in social technology we can stay together in groups of tribe size, with some higher-hanging fruit in groups of up to city-state size, and so on. When you get to the largest organizations in the world (China and the Catholic Church, if you’re wondering), you have to pull out all the stops – a pyramid of subdivisions of subdivisions of subdivisions, each choosing representatives to the next-highest level which then chooses its own representatives and so on until you reach one guy at the top. He suggested that if my old micronation or the Bay Area rationalist community want to get past their growth limitations, they should try something similar – which is good advice, except that in fifteen years of trying to simulate a country it did occur to us to have provinces and that didn’t seem to help much.

I am not sure how much I buy this theory. Having a systematizing disposition, it is really tempting for me to start carving up reality at these lines – startups with < 12 people, small businesses with < 150, et cetera - or to start noting that most of the college clubs I've seen have like a dozen core members, no matter how big the college is. But I find myself engaging in certain thought patterns that I have tagged as "trying too hard to fit data to a theory". At University College Cork the debating club often got twenty or thirty people; and if someone told me the natural way to slice businesess was <5 people, <30 people, et cetera, I could kind of see that too. And also there are so many other factors - whether a group is online vs. meatspace, whether it's living together versus meeting once every few weeks, whether it's strongly enforced (like a company or military unit) or voluntary (like a social club), whether we expect its members to also have other relationships somewhere else (members of MMORPG guilds presumably have real-life friends and coworkers as well, but members of hunter-gatherer tribes might know literally nobody except their cotribesmen). Even if there were interesting limits to the size of cohesive social groups, I would expect these factors to make them different in different domains - the size restrictions governing the growth of online micronations should be different from the size restrictions governing the growth of startups, even though in practice they're often around the same size. But I do notice that the Bay Area rationalist community is probably around 150 people and having trouble growing, and the size of my micronation was around 12 people and had trouble growing. And that after ten years in micronations, the rule that no matter what happened we would never get more than a dozen or so people seemed like a law of nature. And that I just checked after two years outside the community, and it looks like one new person came in to replace me and no one else, and they’re still stuck around twelve. So there does seem to be this very real community failure mode where everyone is socially saturated with each other and new people feel like they can’t break in. I’m less confident it always happens at a specific number or set of numbers across all domains of community – but who knows?