With social media platforms like Facbeook, Instagram and Twitter learning our browsing patterns, it becomes easier to neglect information we normally don't care about. Although practical, this topic has a created a lot of controversy recently since it can be considered an indirect form of bias for anyone absorbing information through Facebook.
Recent posts from sources where the majority of shared articles aligned “very liberal” (blue, on the left) and “very conservative” (red, on the right) in a large Facebook study.
In 2015, the journal Science published a research paper by Facebook scientists (Bakshy, Eytan; Messing, Solomon; Adamic, Lada, 2015, “Replication Data for: Exposure to Ideologically Diverse News and Opinion on Facebook”, http://dx.doi.org/10.7910/DVN/LDJ7MS, Harvard Dataverse, V2) which looked at how a subset of the social network’s users reacted to the news appearing in their feeds.
For six months, Facebook tracked and analyzed the content shared by 10.1 million of its users (who were anonymized). These users had identified their political views in their own profiles on Facebook. Analyzing these users’ political labels, the researchers categorized each as very liberal, liberal, neutral, conservative or very conservative. The researchers also produced a list of the top 500 sources of content tracked in the study.
For a site appearing in the Journal’s red feed, a majority of the articles shared from it were classified in the study as “very conservatively aligned.” How did Facebook researchers calculate that? They examined the self-described political leanings of people who shared links on Facebook to calculate a political “alignment score” for each link. Using that score, they grouped the links into five categories, from “very conservative” to “very liberal.” So that site’s current content appears in the Journal’s conservative column because more than half of its links fell into the “very conservative” category during the period of the study (July 7, 2014 through January 7, 2015).
To appear in the Journal’s blue and red feeds, posts must have at least 100 shares, and come from sources with at least 100,000 followers.
The Journal removed sources that were no longer active and others, such as Twitter and YouTube, which were simply mechanisms for sharing a wide variety of content. The reason you won’t find The Wall Street Journal or any of its biggest competitors in either feed is that their content was shared by Facebook users more broadly across the political spectrum.
Are you saying these sources are conservative and liberal?
No. If a news source appears in the left column, its links were frequently shared by Facebook users whom researchers classified as “very liberal,” based on self-described political leanings. In the right column are sources whose content was widely shared by Facebook users identified as “very conservative.”
Are you choosing which post to include on each side?
No. The content is being pulled automatically from these sites’ public Facebook pages, once an hour using Facebook’s software developer tools (specifically, the Graph API).
Are you saying that these views actually simulate what a conservative or liberal sees?
No. It’s possible that users have a wide variety of news items appearing in their feeds. These are simply posts from sources that aligned with a majority of users of a particular political view in Facebook’s study.
Why did you build this? Aren’t you just reinforcing assumptions about the other side’s biases?
We built this presentation because it’s hard to see these opposing views side by side. Facebook users who are curious about opposing viewpoints may be apprehensive about recording a like for a particular news source—an action which may be seen by other friends. (You can make likes private in Facebook user settings.) This tool gives people anywhere on the political spectrum the ability to see current discussions about newsworthy topics from both very conservative and very liberal viewpoints.