The News Feed is probably Facebook’s most important asset, and there’s no doubt the company spends a ton of money researching algorithms to feed you optimal content. But it turns out algorithms may not be enough – sometimes asking real people direct questions yields better results.
According to Backchannel, since News Feed’s inception in 2006, Facebook has realized that simply observing reactions to a user’s feed and counting likes and shares does not fully correlate with whether people are enjoying the content they’re fed.
A News Feed Project Manager made the analogy of feeding people donuts. If you bring me donuts, I’ll almost always certainly eat them. Donuts are tasty. But that doesn’t mean I wouldn’t rather eat something else instead.
To improve News Feed, Facebook has a random selection of testers around the United States. Using a special version of the social network, the raters are asked to interact with the content normally, then answer questions about their experience and write a paragraph describing their feelings regarding specific stories.
Although this branch of research only began in August of last year, it’s already yielded some interesting insights, including corroborating that the most relevant Facebook content is almost always vital posts and images from close friends.
Additionally, the types of “like” you give to your friends’ posts often are simply meant to convey your connection with the author, rather than any real interest in the story – behavior that is different with regular news stories. Commercial content, on the other hand, still does not relate to user’s interests as much as the company would like.
Facebook currently has about 600 testers on the project, but intends to continue adding testers as the project continues – eventually hoping to reach all of its users. With a continued trend towards automation, research done by interacting with actual people is a welcome change.
➤ How 30 Random People in Knoxville May Change Your News Feed [Backchannel]