Wednesday, May 13, 2015

Beyond Facebook’s Impact on Political Polarization in the U.S.

Any time “scientists” at a company purport to have done a study involving said company in any way, the public has good reason to be suspicious of the reported conclusions. Were the folks running the company really intent on providing credible information, they would use independent scholars (i.e., not being compensated by the company). Such a management would want to obviate even the appearance of a conflict of interest—their desire to provide the public with an answer being so strong. So the management at Facebook may not have been very invested in providing the public an answer to the question: how much influence do users actually have over the content in their feeds? In May 2015, three “Facebook data scientists” published a peer-reviewed study in Science Magazine on how often Facebook users had been “exposed to political views different from their own.”[1] The “scientists” concluded that if users “mostly see news and updates from friends who support their own political ideology, it’s primarily because of their own choices—not the company’s algorithm.”[2] Academic scholars criticized the study’s methodology and cautioned that the risk of polarized “echo chambers” on Facebook was nonetheless significant.[3] I was in academia long enough to know that methodological criticism by more than one scholar is enough to put an empirical study’s findings in doubt. Nowadays, I am more oriented to the broader implications of the “echo-chamber” criticism.

Although the study’s primary question concerned how much what users see on their feeds is due to the company’s algorithm, the question of whether Facebook was a contributing factor in the increasing political polarization in the U.S. was at issue. “What we do show, very definitively,” Eytan Bakshy, who worked on the study, said, “is that individuals do have diverse social networks. . . . The majority of people do have more friends from the other side than many have speculated. We’re putting facts behind this.”[4] Facebook users who self-report a liberal or conservative orientation had, on average, 23% of their friends with an opposing political ideology; and 28.5% of the hard news that such users encountered on the News Feed cut across ideological lines, on average.[5] We know from neuroscience that the human brain privileges people who are similar, so these low percentages do not necessarily mean that Facebook’s algorithm has a contributing impact. That it could be is perhaps of more significance.

According to the 2014 Political Polarization in the American Public study by the Pew Research Center, political polarization increased from 1994 to 2014; the emptying out of a “middle” is especially pronounced among the politically engaged.[6] The proliferation of news networks specializing on particular political market-segments and the ability of social-network users to pick what they encounter are arguably contributing factors, and subtle leanings from a giant company’s algorithm could play a significant role too on an aggregated scale.

The potential in terms of political influence on a mass scale with a usership as large as Facebook’s warrants consideration, however. At the time of the study, Mark Zuckerberg, founder and CEO of Facebook, had the wherewithal to subtly push a political point by carefully amending the company’s algorithm (as well as by using his public platform). Imagine Starbucks' Howard Schultz, who had felt free to use his company (and its employees) to further political issues (and positions) he values, in Zuckerberg's position at Facebook. Would users appreciate being manipulated, subtly or not, into discussions on race that tilt to Schultz's position?[7] The impact on public policy could be astounding simply on account of the proportion of Americans who actively use Facebook. The mining of user-data would give such a CEO even more power not only over particular users, but also on a societal level.[8] The problem, in other words, is that of an unelected person having such massive political power in a viable representative democracy. Similar to the rationale of anti-trust laws, limits on the size of social-media companies in terms of usership could therefore be justified with some degree of precedent.

See also "Taking the Face Off Facebook."




1. Alexander B. Howard, “Facebook Study Says Users Control What They See, But Critics Disagree,” The Huffington Post, May 12, 2015.
2. Ibid. I put the quotes around “scientists” to make the point that the conflict of interest renders the label itself controversial in being applied to the study’s investigators.
3. See, for example, Christian Sandvig, “The Facebook ‘It’s Not Our Fault’ Study,” Multicast, Harvard Law School Blogs, May 7, 2015.
4. Howard, “Facebook Study.” See Eytan Bakshy, Solomon Messing, and Lada Adamic, “Exposure to Diverse Information on Facebook,” Facebook Blog, May 7, 2015.
5. Ibid.
6. Ibid. See “Political Polarization in the American Public,” U.S. Politics & Policy, Pew Research Center, June 12, 2104.