Nieman Foundation at Harvard
HOME
          
LATEST STORY
Stat is publishing a print section in Sunday’s Boston Globe — and it might be coming to a paper near you
ABOUT                    SUBSCRIBE
Sept. 1, 2016, 11:18 a.m.
Aggregation & Discovery
LINK: qz.com  ➚   |   Posted by: Joseph Lichterman   |   September 1, 2016

So how does Facebook’s Trending algorithm actually work? Very few, if any, people outside of 1 Hacker Way really know, but over at Quartz, reporters Dave Gershgorn and Mike Murphy have tried to piece it together.

Facebook last week said it was changing its Trending feature by eliminating the short sentences that described each story. Instead, Trending now just features broad topics surfaced by the algorithm. (On Thursday morning, Van Morrison — it’s his birthday — and Brown Eyed Girls — the South Korean band, not the Van Morrison song — were trending on my Facebook.)

Facebook recently laid off 26 contract editors and curators, with a team of engineers now overseeing the algorithm. And hours after the company made the switch, it had already run into trouble by promoting fake news stories and viral stories, such as a man engaging with a McChicken sandwich in a lewd fashion.

According to Facebook’s guidelines, the engineers overseeing Trending are “responsible for accepting all algorithmically detected topics that reflect real-world events.” Quartz noted that the document doesn’t specifically mention hate speech, only instructing the engineers to delete stories that are sexual in nature or that aren’t tied to real-world events.

Examining previous news reports and Facebook’s own patents, Quartz found that “mentions” are a critical component to the algorithm. In a July 5 patent awarded to the company, it defines mentions as specific terms or hashtags that are trending:

According to the patent, as the algorithm processes mentions and finds matching keywords, like NBA Finals or Super Bowl, it begins to sort them into a topic. Then the topic is given a Trending score based on geographic location of users posting about it, the timeframe during which those posts were made, and weighted number of mentions, according to the patent. The most important part of this score seems to be timing—the more mentions over a shorter time period, the higher the Trending score the topic gets. After the Trending score is assigned internally, and ostensibly approved by a reviewer, the Trending topics that are displayed are personalized for each user. Geographic location is a big factor, but the algorithm described in the patent also weighs “gender, race, national origin, religion, age, marital status, disability, sexual orientation, education level, and socioeconomic status,” as well as previous interactions with other content.

Of course, this is only a patent and doesn’t necessarily correspond to what Facebook is currently using. However, the description in the patent matches well with the information Facebook has publicly released.

Facebook also said that it assigns less value to users who post more frequently.

Even though it’s just a patent, the information gives us a window into how Facebook is thinking about the algorithm and what the implications are for news outlets and news consumers. More than 40 percent of Americans get their news from Facebook, according to a Pew report earlier this year. (Mark Zuckerberg also said this week that Facebook is a technology company, not a media company.) While it’s unclear how many of them actually use Trending, which is difficult to find in Facebook’s mobile apps, it clearly still wields some power.

On Twitter, University of North Carolina professor Zeynep Tufekci said Facebook’s (and other social networks’) approaches to trending topics could have serious implications around the world.

Photo of Facebook headquarters by Jimmy Baikovicius used under a Creative Commons license.

Show tags Show comments / Leave a comment
 
Join the 35,000 who get the freshest future-of-journalism news in our daily email.
Stat is publishing a print section in Sunday’s Boston Globe — and it might be coming to a paper near you
The health and life sciences site is in talks with other newspapers about republishing its coverage in print.
A new database of fake news sites details how much fakery has spread from Trump v. Clinton to local news
Plus: The New York Times walks back an extremely popular tweet, California adds media literacy to its curriculum, and the KIND Foundation tries out a “Pop Your Bubble” app that nobody is going to want to use.
Nieman Lab is looking for more stories of digital innovation outside the U.S., and we’d love your help
Have ideas for things we should cover? Want to help us cover them? Want to help us translate stories so that they reach more people? Join our new Slack community!