Narrowcasting and the Facebook News Feed
I call it “Regression to the LOLCat”: when our friends and followers on Facebook are split across high school, college, sports teams, and professional life, it’s tough to find content to post that will satisfy them all. We either develop niches (like posting professional items on Twitter and personal items on Facebook), or post content of general interest. More often than not, those are universally accessible items like LOLCats. How much is your Facebook feed consumed by boring, all-purpose content?
I think this is A Problem. There’s no easy way on Facebook to tell my high school friends that our old choir teacher is retiring without also spamming my grad school cohort via the news feed. The current model is “broadcasting”; the model I want is narrowcasting. Whether the interaction for this is implicit or explicit (through lists?), give me some confidence that I can guide a status update more toward family members than the rest of my social network.
It occurred to me today that Facebook’s routing algorithm, called EdgeRank, does allow you to hack this a bit. Courtesy of TechCrunch, here’s the formula:
I’ll cut through the math using words. Whenever somebody interacts with a news feed item, they create an edge to that item. So if I comment on a friend’s new puppy photos, I’ve created an edge to your photos. When trying whether to show the photos in your news feed, Facebook looks at how closely you interact with everyone who has an edge to the item. So, with the puppy photos, it considers your affinity to the friend who created the photos, and then me because I commented on them.
This all comes down to — initialization matters. If my high school friends are the first to comment on a news feed item, the EdgeRank of that item for other high school friends is high. So, other high school friends will see the item. If grad school friends are the first to comment, then other grad school friends are likewise going to see it.
I’m also interested in using more content-based approaches, like the topic-based approach we use in Eddi, covered in Technology Review. So, if I tweet about HCI, show it to my grad school friends; if I tweet about Woodbridge High School, show it to my high school friends.
Hmmm.
Check out Jyri Engestrom’s ‘Social Objects’ particularly as they relate to the Activity Streams spec ( which Facebook is already using )
http://everwas.com/2009/07/jyri-engestrom-on-social-objects.html
Fascinating topic, Michael.
You might want to include the fact that Facebook does offer per-item privacy settings (as a way of reaching particular sets of people) in future discussion of this — even if you dismiss it as too difficult to use.
@Dean: Great point. Right, if you’re willing to create lists or groups on Facebook, then you can explicitly target them. My personal take is that the start-up cost is too high, but at the same time any automated method might be too error-prone to trust. It’s a touchy issue.
[...] Narrowcasting and the Facebook News Feed An insightful post. I have lot of annoyances with Facebook and the inability to make only a group [...]
Very cool to see the formula posted in the clear. I did not know there was a ranking that took time-to-respond as a factor. Fascinating.
I once experimented with narrowcasting on Facebook setting up a specific group with just my high school classmates to send them targeted messages. Things broke down though because it was impossible for those receiving my messages to know they were only going to a small group so their responses were as if they were responding to all their Facebook friends.
Until a product can enable explicitly private discussion in the same way email or private forums can (you can see everyone on the thread), the tools for narrowcasting will remain woefully 1.0.
Ian
[...] algo is called EdgeRank and he describes it as such: I’ll cut through the math using words. Whenever somebody interacts with a news feed [...]