Statistics, Rarity, & Social Networks
On why thriving social networks are actually based on rarity.
Consider what a bad job traditional social network thinking does about the value of the rare things we have in common.
Traditional wisdom would say that rare things aren’t as relevant as common things, and statistically what is rare is what is excluded or undetectable because low frequency. But cultural or social rarity is always perceived by humans as 1) of primary importance and 2) always as relative to some out-group, some ‘Them’ vs an ‘Us’. For example, we don’t much care that we’re all humans despite the fact that sentient life is one of the rarest things in the Universe. In other words, rarity only binds us if the group is interacting with outsiders for whom the rare thing in common unites us and makes us special.
To the group in and of itself, this shared rarity doesn’t have much salience. In fact, too much rarity in common might repel us as it robs us of our specialness at the individual level. Whereas if I have something rarer than you, I’m going to talk about it.
More than that, anything we might agree on outwardly as something we have in common, such as our race, will be a source of nitpicking one-upmanship as we vie for supremacy over the smaller and smaller differences that we can find when we’re alone. Freud called this phenomenon “the Narcissism of Small Differences”.
The same logic operates with individuals sharing a rare trait. When I am a 49ers fan in SF among the general population, this does not attract me to anyone. But when I bump into a 49ers fan at a soccer party of English people in London, it makes us feel Destined to know each other and we may go out on a date. Mutual rare coincidence relative to an out-group is the basis for the feeling of ‘destiny’ and it’s a cognitive bias rooted in evolutionary psychology. It’s social fission, it’s speciation, it’s what drives all culture, it’s the eternal struggle to be special and yet also belong.
To sum up, traditional models of predicting social connection are based on similarities without regard to real relevance, because they leave out the statistical rarity of the out-group context that determines which similarities are really relevant (as in the Beduin saying, “me against my brother, my brother an I against my family, my family against my cousins, all of us against the World”). Two 49ers fans means nothing in the SF social scene. It’s Fate in London. Traditional predictive models would pick the least relevant traits, such as us both being American, because they are the most common things that people have in common. But it’s that we both went to that Phish concert at the Glastonbury Festival in 1991, an event that would not register on any AI algorithm as a statistical pattern, that seals our social connection.
