Facebook’s News Feed Knows What You Did Last Summer

When Facebook turned on News Feeds for the first time in September 2006, many users were spooked by the privacy implications of a system that logs and broadcasts your activity to your friends. However, much of the concern quickly subsided – perhaps because it’s worked so well, the system that powers Facebook’s News Feed hasn’t gotten much attention since then. As Facebook engineer Justin Rosenstein remarked this summer,

“News Feed works so well that it’s sometimes easy to forget how sophisticated it is under the covers. If you have a lot of friends, your Facebook home page is displaying only a tiny fraction of what’s going on in your social network. The system that selects the right subset is impressive from both an AI perspective (with a ranking algorithm that uses signals based on user behavior throughout the site) and a systems perspective (efficiently processing 1.2 trillion story candidates every day).”

According to Facebook, News Feed publishes just a little more than 0.2% of the stories it considers. This means that of the 1.2 trillion story candidates Facebook considers every day, only 2.4 billion get published in users’ News Feeds. And on a per user basis, that means each user sees only 60 out of about 30,000 possible story candidates on any given day.

The News Feed Algorithm

How does Facebook decide which ones to show? Of course, Facebook will never publish the weighting algorithms, which are constantly changing. But according to Facebook’s Rosenstein, while the weighting algorithms do apply some general principles (see NFO is the new SEO for tips on optimizing your feed items), they primarily rely on behavior specific to each user.

In fact, Facebook considers nearly every available source of data it has on each user to help calibrate their weights and deliver the best stories:

  • Whose profile pages you visit – and how frequently, how recently, and how often
  • Who you message, and who messages you
  • Whose walls you write on, and who writes on yours
  • Who/what you search for
  • What’s on your profile
  • Who you invite to events and groups, who accepts, and who invites you
  • Who you tag in photos, and who tags you
  • Which News Feed items you’ve clicked on before

Facebook is able to choose the right 60 News Feed stories for you because, based on your behavior, Facebook knows who and what is important to you.

This is a Big Deal – the News Feed has quietly revolutionized the way information flows across the social graph. The quality at which it subtly functions is truly a landmark technical achievement, and Facebook users vote with their feet: 50% log in every day.

Privacy & Security

Of course, with this amount of personal information, there is always a necessary privacy angle. According to Facebook’s privacy policy, Facebook reserves the right to log everything users do on the site in order to offer “personalized features:”

When you use Facebook, you may set up your personal profile, form relationships, send messages, perform searches and queries, form groups, set up events, add applications, and transmit information through various channels. We collect this information so that we can provide you the service and offer personalized features.

While Facebook offers extensive privacy controls, it doesn’t give users the option to opt out of data collection for personalized features. This means that if you’re a Facebook user, Facebook is keeping tabs on everything you do whether you know it or not. As long as this information is always used properly, Facebook will increasingly be able to provide better features like the News Feed.

However, improper use or distribution of this information could be devastating. (How embarrassing would it be if that crush or co-worker found out how frequently you were viewing their profile?) Facebook has never allowed this to happen on the site (and says it never will). In addition, Facebook has not allowed third party applications (like Trakzor) to log or display this information either (this is possible on MySpace).

However, there were recent rumors that Facebook employees may have abused access to this personal information by emailing personal friends who they discovered had visited their profile page frequently. As Facebook hopes to occupy a trusted space in the lives of tens of millions of people, Facebook needs to establish its trustworthiness on these matters and quickly clean up its act.

Future of News Feed

Ultimately, the News Feed has even more potential to become the Most Important Way Information Spreads Online. Most pressing is the need for News Feed to publish application stories to friends who don’t have the given application installed.

Currently, only the News Feed story that says you’ve added a given application is visible to all of your friends (and it alone generates a meaningful number of installations):

However, once you’ve added the application, stories about your interaction with the app are only visible to your friends who also have the app. While intended to reduce the incentive to News Feed spam, this restriction cripples the News Feed’s ability to help highly engaging applications spread. If Facebook wants to encourage the development of applications that are deeply engaging over long periods of time instead of fly-by-night gimmicks, they should allow these stories to be visible to all your friends.

As other social networking services open their own platforms over the coming year, they will be hard pressed to close the competitive gap Facebook has created with News Feed soon. Within six to nine months, News Feed will have considered over one quadrillion News Feed items on which to tune its algorithms since its launch over a year ago. I just wish News Feed would let me see more of my 30,000 stories a day.

[tags]facebook, news feed, virality, privacy[/tags]