Race and Social Network Sites: Putting Facebook’s Data in Context

by Guest Contributor danah boyd, originally published at apophenia

A few weeks ago, Facebook’s data team released a set of data addressing a simple but complex question: How Diverse is Facebook? Given my own work over the last two years concerning the intersection of race/ethnicity/class and social network sites, I feel the need to respond. And, with pleasure, I’m going to respond by sharing a draft of a new paper.

But first, I want to begin by thanking the Facebook data team for actually making this data available for public dialogue. Far too few companies are willing to share their internal analyses, especially about topics that make people uncomfortable. I was disappointed that so many academics immediately began critiquing Facebook rather than appreciating the glimpse that we get into the data they get to see. So thank you Facebook data team!

There are many different ways to collect quantitative data involving categories like race, ethnicity, class, gender, sexuality, etc. None of them are perfect. Even asking people to self-identify can be fraught, especially when someone is asked to place themselves into a box. Ask a self-identified queer boi to identity into the binaries of “female/male” and “gay/straight” and you’ll see nothing short of explosive anger. Race certainly isn’t any prettier, let alone ethnicity or class. The salience of these qualities also depends on what we’re trying to measure, what we’re trying to say. For example, if we’re talking about people who experience being targets of racism, should we concern ourselves more with self-identification or external labeling? At the coarsest level, we often assume race to boil down to skin color, meaning that we have to take into account how people read race, how they experience race, how they identify with race. We must always remember that race is a social construct and one’s experiences of race are shaped by how one perceives themselves in relation to others and how others perceive them. And the very notion of race differs across the globe.

Of course, this is bloody messy. And ethnicity and class are even harder to locate because self-identification isn’t always the best measure. Heck, while Americans have learned to self-identify with race (thanks to countless forms), we aren’t typically asked to self-identify with ethnicity or class. So these are pretty murky territories. As a result, scholars and demographers and marketers and many others have different ways of trying to measure these categories. None are perfect. We can debate endlessly about which is better but, personally, I think that does the conversation a disservice.

In trying to measure race (and, partially, ethnicity) of its users without having self-identification, Facebook decided to use a statistical technique known as mixture-modeling to make a best guess as to the racial makeup of its user base. They go to great lengths explaining what they did, but it is this graph that we should be attentive to:

Next page