Facebook on Tuesday rolled out new changes to its algorithms that determine what posts users of the social network see in their feeds. The new changes increase the scrutiny of posts linking outside Facebook and reduce the visibility of posts with “clickbait” headlines by pushing them further down users’ news feed page.
Last August, Facebook cracked down on specific news sources that consistently posted clickbait headlines, or “headlines that intentionally leave out crucial information, or mislead people,” Facebook’s Alex Peysakhovich and Kristin Hendrix explained at the time. This approach looked broadly at a publisher’s posts to Facebook and lowered their relevance ranking if the publisher consistently used headlines with phrases like “You won’t believe what happened next” and “Her reaction was priceless!”
The latest change increases the scrutiny given to news posts by altering the way they’re assessed for “authenticity.” This doesn’t mean that people won’t still find and click on the bait or see fake news; they’ll just have to scroll a little further down the page to find it first.
In an announcement posted yesterday, Facebook engineers Arun Babu, Annie Liu, and Jordan Zhang explained:
- First, we are now taking into account clickbait at the individual post level in addition to the domain and Page level, in order to more precisely reduce clickbait headlines.
- Second, in order to make this more effective, we are dividing our efforts into two separate signals—so we will now look at whether a headline withholds information or if it exaggerates information separately.
- Third, we are starting to test this work in additional languages.
The “signal” splitting means that Facebook’s algorithms assess headlines in two separate steps: one set of algorithms is optimized for looking for language typically used to hide information, while another set assesses posts for language associated with exaggeration of the details. The algorithms were trained on hundreds of headlines identified by Facebook researchers and engineers as deceptive
A hit from either type of check will lower the post’s priority and bury it lower in users’ news feeds, but it won’t remove the post entirely. This is consistent with Facebook’s efforts to fight “fake news” by highlighting posts with warning labels that caution readers about the news source and reduce their visibility by adjusting their ranking.
Of course, these algorithms will likely be defeated quickly by clickbait producers—once they learn a new set of phrases for their headlines. And algorithms still can’t fact check stories. At least, not yet.
More Info: arstechnica.com