Tony Sheng     Projects     Contact

Segregation by social networks

Schelling’s model of residential segregation

My brother graduated from Duke this past weekend (yay Gary!). He whirled around campus exchanging warm embraces, hosted parties for his closest friends, and rejected sleep for several consecutive days to soak in every last remaining minute with his college brothers and sisters. The value of his experience was not symbolized by a leather-bound diploma, but by the texts and hugs and promises exchanged with the people to whom he felt eternally bonded.

Meanwhile, my parents and I sat through several long speeches by old white men who wanted in these final hours to remind us why the experience at Duke was so valuable to our loved one. A few consistent themes emerged over the course of the weekend, but the one that stuck out was the value of empathizing with those that are not like you and the practice of channeling that empathy into material positive impact in the real world.

So for the students, the experience of graduating is celebrating all of these people that are deeply like you and will be--as far as anybody can tell at this point in time--more like you than anybody ever will be. In contrast, the administration’s parting words--their mission--is for the students to understand people not like you and make the world better for them.

One might argue that the tribalism of a college education can’t be avoided, and that the affinity one feels for their college friends isn’t at odds with the aspiration of serving those not like you. One might continue to argue that the demographics of a graduating class are more diverse than any other group a student could join. One might further argue that only by instilling these shared values over the course of a four-year education can classmates band together to work together and create positive change for those unlike them.

My experience tells me that these arguments all have merit, but fail to ensure the success of the mission--not because of any active resistance, but simply because of our natural behavior and the impact of technology on our social organization. We’d like to see a trend of increasing exposure to people not like us following graduation. Unfortunately, the social networks of “elite students” tend to grow more homogenous over time.

People naturally like to organize in groups with people like themselves. This makes sense from a basic evolutionary perspective. When we perceive people as like us, we assume they have more genes in common with us so we feel a natural affinity. For an example of how this plays out, check out Schelling’s model for residential segregation. In a nutshell, his model suggests that even the tiniest preference for living next to similar people leads to highly segregated neighborhoods.

Check out this article to learn more about Schelling’s model and play around with a simulator.

Geography and employment already create environments where it’s very easy to maintain tribes of people like you. Most college have a few major cities where alums start their careers. For Duke and Amherst, it was very obviously New York City. The UC schools in California congregate in Los Angeles and San Francisco.

Likewise, most schools have a few clear job paths that students typically pursue. Duke and most other elite schools trend towards consulting, finance, and graduate school (specifically MD and JD).

So now we have a preference for alums of our school in our geography in our specific profession. Instead of surrounding ourselves with people that are not like us, we’re surrounded by people that are very much like us.

Social media has an interesting impact on all of this. When LinkedIn first came out, it was positioned as this remarkable tool to connect you with people that were unlike you through your existing social graph. But we’re realizing now that social media doesn’t provide much breadth to your social circle, it provides depth. One might imagine alumni groups named “Duke Finance Professionals in New York City.” Here are three concentric circles that make you more and more closely tied to people like you. As a result, the people you meet through these connections will more often than not be much more like you than unlike you.

Importantly, this kind of social network segregation doesn’t happen because of an active attachment to in- and out-groups (xenophobia). In fact, Schelling’s model fails to result in segregation if there are very high preferences for their own group. That is, a racist community never settles into segregated neighborhoods because they’re always moving around because they’re always unhappy about how close they are to the other group.

So the argument that college makes you more tolerant and accepting of groups not like you actually enables this kind of subconscious social segregation. Intolerance would make us more aware of people unlike us. A slight, natural preference for people like us leads to blindness. We don’t even notice that we’ve settled into homogenous social networks.

I don’t think this is inherently a bad thing. Friend and professional groups are naturally more like each other than not. That’s why they feel comfortable with each other and like spending time together. I just don’t know where the desired outcome--empathizing with those unlike you and making positive change for these people--comes into play.

Given trends in information consumption, our awareness of people unlike us and the problems they face is probably diminishing. The content we see on our Facebook feeds is governed primarily by what our friends like to click. How are we going to see what matters to people that are not like us if we’re delivered content based on what matters to people like us? If mainstream media is bisected by conservative (FOX) and liberal feeds (everybody else?) and we’re worried that this creates an echo chamber where people only hear what they want to hear, imagine a world where we can get infinitely more specific than a bisection. Imagine machine-learned echo chambers of narrower and narrower slices.

On the other side of the spectrum, apps like Yik Yak reward you if you post or engage with content that your particular community endorses. Share something that’s popular, get a lot of upvotes. Passive users of the app start to understand what they “should” believe by osmosis, passively observing activity in their Yik Yak community.

So what’s the solution? It’s probably not as easy as Facebook friending a bunch of people you didn’t go to school with.

The dean of the computer science program’s parting words to Duke grads was this: we’re really good at making technologies that connect people that are like each other, but we’re really bad at making technologies that connect us with people that are not like us... my hope is that you will go off and help solve this problem.

We accept this sentiment as objectively good, but my worry is that left to our own devices, we end up never tackling this problem. It’s not our natural preference. Importantly, today’s social media is not the root cause of segregation, it just exacerbates the default conditions of modern humanity. But you have to wonder whether new forms of social media can change the conditions.

I don’t know the answer here because I recognize this as a problem of my own. My social networks are laughably specific. So I’d like to open this up for discussion. What’s your take? What needs to be done? What have you seen work?

Thanks to Gary Sheng for reading an early draft of this post

If you enjoyed this post and want more like it, join thousands of readers on the weekly newsletter.