Facebook Is Making News Feed Better By Asking Real People Direct Questions


Social Media Life

Getty Images



It’s a well-known fact that Facebook’s flagship feature, News Feed, is run by algorithms.

Essentially, invisible computations are going on all the time that automatically optimize future items you see on your feed, depending on the actions you take now—what you click on, what you like, what you comment on. The goal, as CEO Mark Zuckerberg told WIRED in 2013, is “to build the perfect personalized newspaper for 1.1 billion people and counting.”


But Facebook knows that it can do better than relying solely on these cold computations.


As detailed in a new piece on Backchannel by former WIRED writer Steven Levy, Facebook is currently running a focus-group-like program that asks people direct questions about News Feed items in an effort to improve post relevance. According to Levy, the pilot program started last August, testing just 30 Facebook users in an office in Knoxville, Tennessee.


It has now expanded to 600 people around the country, who are paid by Facebook to work answering News Feed questions four hours a day from home. Eventually, Facebook could offer some kind of direct questioning to its entire population of users.


The project works like this: each of these 600 Facebook users is presented with 30 top News Feed stories in a random order. Then they go through each story one by one. They can comment, share, follow a link, or choose to ignore the story. After that they answer eight questions about each item, including how much they cared about the subject of the story, how welcome the story was in their News Feed, how entertaining it was, and how much the story connected them to friends and family. Finally, they are asked to write a few sentences describing their overall feelings about the News Feed story.


Facebook itself acknowledges there are problems with how News Feed is currently set up. It’s already very good at delivering personal news from close friends—things like marriages, childbirths and vacations—but it’s also overrun with items that are sugary sweet and designed to tug at your emotions, which Levy has dubbed the “Dozen Doughnuts problem.”


The donut-y content contrasts with a “vegetables” of real journalism and hard news. When so many of those donuts are presented to you at a time, you’re bound to click on at least one item. And that click sends a strong signal to Facebook: you want to see more of the same thing.


Facebook could interfere. But especially in the case of News Feed, it prefers not to be heavy-handed. “We really try to not express any editorial judgment,” Adam Mosseri, News Feed product director, tells Levy. “We might think that Ferguson is more important than the Ice Bucket Challenge but we don’t think we should be forcing people to eat their vegetables even though we may or may not think vegetables are healthy.”


Preliminary results have already emerged. As expected, news from close friends—especially tagged and photo stories—has been consistently rated as highly relevant. But other things, like the meaning of a “like,” has proven to be more ambiguous. It could mean anything from the approval of a story to validation of a user’s connection to the author.


Unfortunately, so far, it looks like users are less willing to engage with “meaningful” stories or news, preferring anything that triggers a strong emotional response. But Facebook is hopeful that when it begins asking users about sets of stories instead of individual items people will start to reward informative content.


Though some Facebook employees are quoted in Levy’s story as wanting to do the right thing by fixing the News Feed, the real reason why Facebook may have a vested interest in making News Feed the best product it can be is glossed over. Facebook made $2 billion in ad revenue last quarter, more than two-thirds of its total $3.59 billion in ad revenue for 2014.


And where do those ads live? In News Feed. If the social network can crack the problem of what users really want from News Feed, they can presumably apply those learnings to ads, too—and make those ads irresistible to its users in the process.



No comments:

Post a Comment