For a long time, the internet hasn't been just one thing.
Read more: Small talk should be banned
Because our every move within and across sites and apps is tracked, we now receive bespoke news, advertisements and search engine results. This is either celebrated as good - "I can finally get what I want without having to wade through endless crap" - or condemned as creepy; corporate surveillance and the end of privacy. Recently, though, personalisation is being blamed for the increasing polarisation of society. Although personalisation certainly plays a role in the rise of echo chambers, it's not the whole story. In fact, it can't be: if we were all being treated like the special snowflakes we are, no kind of community would be possible. Whether we like it or not, polarisation is about communities, just segregated ones: from groups that spread the same fake news (while also accusing others of doing so) to the rise of extreme political groups. We are segregated because networks put us into virtually gated communities.
So, your recommendations don't just depend on your actions. If they did, they would be pretty limited in scope. You really don't matter; what matters is what people "like you" do. Thus, it's easy to predict your age and political ideology based on the behaviour of people who follow you on Twitter, especially if you follow them in turn. The Chicago Police created a "heat list" - people most likely to murder or be murdered based on social-network analysis. But again, you don't need to know the people in your virtual neighbourhoods. They can live many kilometres away from you: you both like Duck Dynasty and hate Game of Thrones; you both like country music and hate rap; you both drive similar cars. We're sorted into neighbourhoods based on our likes and dislikes in order to uncover hidden relations and to foster new ones.
The driving logic here is homophily: the idea that people who are alike act alike, that birds of a feather naturally flock together, that similarity breeds connection. Homophily makes networks by creating clusters, so you're not just linked to others because of your IP address or geographic location. This means you're easier to find and it's easier to find others. This principle is allegedly as old as Plato and Aristotle, who declared that similarity drives friendship and love. Not surprisingly, race and ethnicity, gender and class are all considered to be homophilous - that is, both markers of great cohesion and division.
Perhaps, but homophily - the creation of network neighbourhoods based in online similarities - assumes neighbourhoods should be and are segregated. Not surprisingly, segregated neighbourhoods are offered as evidence of the so-called naturalness of homophily. But this logic is bizarrely circular, especially since the very term of homophily emerged from a sociological study of US residential segregation. Writing in the 50s, Paul Lazarsfeld and Robert K Merton coined the term "homophily" in their study of friendship patterns within two towns: "Craftown, a project of some 700 families in New Jersey, and Hilltown, a bi-racial, low-rent project of about 800 families in west Pennsylvania." Crucially, they didn't just coin the term, they also created the term "heterophily", and they didn't find homophily to be "naturally" present.
Rather, homophily in their much-cited but clearly unread chapter is one type of friendship formation. Why is this important? Not only does it show the deep ties between the notion of homophily and segregation, it also shows how current uses of homophily have transformed it from one possible form of friendship formation to the default. As homophily has moved from a problem to be solved to an answer to be used, we've become more and more virtually segregated. Network analysis changes the reality it models: Software used by some US courts to predict recidivism - and thus determine sentencing and parole - has been shown to be biased against African Americans. Racial profiling? No shit, Sherlock.
The reasons for the "naturalness" of homophily are usually assumed rather than given. More often than not, it's assumed that people are more comfortable when they're around people like themselves. It's also assumed to be natural. Just as birds of a feather flock together, so do humans. But unlike birds, we're all of the same species. Also, homophily has real problems explaining things such as heterosexuality: if gender is a strong marker of homophily, how do boys and girls ever get together? But the most damning example is segregation itself: segregation happened in the US because of legislation and institutional and economic discrimination. To call segregation natural is to erase great swathes of US history, from slavery to redlining.
So what can we do? As many others, such as Cathy O'Neill and Eli Pariser, have pointed out, this doesn't have to be the end of the story. Homophily might be good at strengthening already existing ties, but it also misses many other things: from purchases inspired by strangers to opposites attracting. Let's start building models that take heterophily – the equally age-old truism as opposites attract – as the basis for connection. Even better, let's build networks based on mutual indifference. This, after all, keeps cities and communities going. Most broadly: let's build search and recommendation engines that don't narrow our perspective, but open us up to the world around us.
This article was originally published by WIRED UK
