Click image to enlarge
As seen in my Facebook social graph, my online relationships tend to be clustered more by geography than shared interests.
To clarify, these shared interests would include events or communities, since all of this can take place in either virtual (e.g. hobbyist forum) or physical (e.g. community center) locations.
In the 90s, Barry Wellman did a related study on one of the world’s first “always-on” Internet suburbs called “Netville“. During his early visit to UB, I recalled him relating an irony of how we communicate predominantly with our physical neighbors, despite us being afforded the ability to base our communication on mutual interests with anyone in the world.
I would surmise that physical proximity still has a higher significance on our communication due to the increased potential impact (threat) it could have on our well being (survival).
See if this is true for you. Try generating your own social graph, then label your clusters. The denser cluster should be ones that are geographically centered. If you can, upload your screenshot to Flickr, then tag it: facebookclusters
Okay so this might not be the most valid test, but it gets us thinking. Chris Lott pointed out how…
- The choice of Facebook for experiment skews the results, making them narrowly applicable to FB not one’s actual social network
- not to mention the obviousness of geographical proximity as a major force in friendships which reflect on SNs…
I would half-agree.
For the first point on the disassociation of online vs. real-life networks, I believe that given sufficient friend connections on Facebook, it would still serve as a decent sample/proxy of our real-life relationships. I did consider factors such as the digital divide, but with an increasingly broader demographic of Facebook participants, this might not impact the test as much as we might imagine. Having chatted/interviewed with a small number of new and senior Facebook users, most are amazed at how many of their friends are already there.
On the second point of obviousness of geography as cause for friendship, I relate back to the idea of the test: To see if the Internet truly encourages us to communicate (including relationship formation) on shared interests regardless of proximity. In essence, the online space would allow for geographic friendships to compete with shared interest friendships, yet geographic ones still appear as denser clusters.
A caveat for using Facebook to test how we center our communication online, would be that friending on social networking services are single-action triggers, and are unrepresentative of longer-term communication. For a more accurate test, I’d need to be able to measure who we tend to talk to (nodes), how often we do (frequency), and where that person resides (location).
I’m still trying to find a better way to conduct this Facebook test, so lets consider this a pilot. Now if we take that the average no. of Facebook friendships to be about 164, then a friendship corpus of about 200 or more should suffice for this test.
Do note that I find friendship counting irrelevant, because our current architecture of social networking services naturally grows our connections. That is, it’s easier to make connections, yet more work to break them. We’re never going to stop meeting new people throughout our lives. Interestingly, while I consider myself an active friender online (currently 620 FB friends) , my geocentric network clusters still hold true!
Here are submissions from my friends…
Joe Hsu / @jhsu / 491 Facebook friends
Denice Szafra / @denidzo / 123 Facebook friends
@denidzo said “yes, but while problematic, it does indicate that I don’t randomly friend people- that I mostly talk to people I already know.”
November Tan / @micamonkey / 836 Facebook friends / 4yrs of Facebook use
Among her thoughts, November said “I find it interesting that my family network runs in parallel clusters. One for each side of the family!”
Jeremy Foo / @echoz / 308 Facebook friends / 2yrs of Facebook use
Jeremy said “I would think that my clusters are based upon events in life rather than location. Its more often than not a classification system that is familiar to you.”
I did consider whether the classification of shared interest and location was arbitrary, since both could be mutually inclusive. An event would be an example of a situation where both coincide. However, since shared interest could exist in physical and virtual place, it’s still fair game.
As iffy as this sounds, I’ll need to compare more social graphs out there, so do contribute your annotated screenshots:
1. Generate your own social graph.
2. Label your clusters by shared interest and location.
3. Upload your screenshot to Flickr, then tag it: facebookclusters.
4. Include your friend count and how long you’ve been using Facebook.
Finally, let me know how you’d improve the test. Also let me know if you’ve found any network tool that lets me get at the data points I’ve mentioned. Thanks!