Friday, December 17, 2010

Facebook friendships and urbanization

Last Tuesday Paul Butler, an intern on Facebook's data infrastructure engineering team, posted a map visualizing the "locality of friendship" on Facebook. Butler used data from friendships between the 500 million Facebook users to create a stunning visualization of the world's largest social network. The colour of the blue lines in represent friendship connections between cities and the white dots are the cities themselves (literally glowing from the density of the friendships within the city).

As interesting as the brightly shining cities and interconnecting lines are (think for a minute about what the lines represent!), many of us found the dark areas to be even more interesting. As is often the case, the most interesting questions came not from what we could see in the data set, but from what we couldn't see.

To that end, Thorsten Gätz produced a beautiful map where the Facebook friends map was put on top of a world map where areas with a high population density were marked in red. He described it as "[a] map which puts the connections of facebook users into context and shows where other social networks have the upper hand."

I am however not sure that population density is the right variable when trying to put the Facebook map into context. Parts of the third world are densely populated but not particularly urbanized. Urbanization should be highly correlated with social network usage, so perhaps that would be a more natural variable to look for in a comparison.

Ten years ago, NASA used data from meteorological satellites to produce a picture of Earth's city lights. Brighter areas are more urbanized, so this map should be useful when we try to put the Facebook map into context. Luckily, NASA's map uses the same projection as the Facebook map, so we can easily superimpose them on each other.

My graphical software skills are sadly lacking when compared to those of Gätz, but after trying my best I came up with this picture:

The red (and glowing) dots and lines are from the Facebook map whereas the blue parts come from NASA's map.

The conclusions that we can draw from this map differs somewhat from those drawn from Gätz's map. The most urbanized regions of Africa seems to coincide with the African regions on the Facebook map. Northeast Brazil, Russia and parts of eastern Europe, the middle east and China are either underrepresented or missing completely in the Facebook map. South Korea and Japan are clearly visible on the Facebook map, but even more so on the city lights map.

The picture that emerges when looking at the Facebook and urbanization maps together coincides with that from a world map of social network that was recently published over at Vincos blog, giving further support to the idea that urbanization is the right context for this map.

Dark areas of maps have always driven mankind to push further and explore the unknown. Today we can explore those dark areas from the comfort of our homes, with a little help of powerful computers and networks of satellites...

Revolutions and FlowingData drew my attention to Butler's Facebook map.

On a side note, it's also interesting to compare these pictures to world connection density and city-to-city router connections map.


  1. Hey, the idea for my map actually stems from a comment on engadget, which called for the same approach you went for, depicted (though not as beautiful as your visualisation) also here:

    But I see now that my flickr description is a bit misleading. As stated in other forums, I wanted to point out the dark areas too. Having no usable datasets on the other networks, I just hinted at those most used in the non-facebook areas :)

    And of course, I wanted to point out that there are still huge dark spots considering internet usage. It may be an obvious thing to state, and my visualisation surely fails to distinct the two reasons for red. As soon as I have time, and the necessary datasets, I will create a better version :)

  2. Looking forward to see what you do when you get more data! :)

    It would be pretty cool to see how the friendship map develops over time. I'm sure that Facebook has similar data recorded over time, which would make it possible to create a map that changes over time, Gapminder style.

  3. Have a look at this map I made. I think this is what you were trying to do:

  4. Yes, that's much better Ian! :)

    On a related note, published a comparison of population density and IP addresses in October: