Is Facebook to blame for people failing to read content that questions their world view? Apparently not, say researchers at the social media giant. They say user choice is the main driving force that skews what someone sees. Does that mean the social network is off the hook? Not quite.
The , published yesterday in the journal Science, dives into the Facebook activity of 10.1 million users in the US, looking at what liberals and conservatives read, shared and engaged with on the platform over six months in 2014. The aim was to explore 鈥渇ilter bubbles鈥� 鈥� the idea, posed by Upworthy CEO Eli Pariser, that online algorithms limit exposure to information that conflicts with someone鈥檚 political leaning.
The authors found that Facebook鈥檚 algorithm had a modest effect on the kind of content people saw, filtering out 5 per cent of news that conflicts with conservative views and 8 per cent for liberals. However, they found that self-screening had a much bigger effect. People showed a clear preference for stories that fit their own world view: liberals clicked on only 7 per cent of conflicting content, while conservatives clicked on 17 per cent.
Advertisement
The authors conclude that individual choices, more than algorithms, limit exposure to diverse content. Or, as Christian Sandvig, an internet policy researcher at Harvard University, put it in a yesterday: 鈥淚t鈥檚 not our fault.鈥�
While the researchers I spoke to highlighted the study鈥檚 merits as a big data analysis, they noted some important limitations.
Not representative
at the University of Maryland points out that it only looked at people who listed their political affiliation on their profile and who fit into the researchers鈥� five-point ideological scale. That turns out to be 4 per cent of users, who are probably not representative of the US as a whole.
鈥淲hat about the other 96 per cent?鈥� he says. 鈥淚t鈥檚 important to get that out there and say that this is based on a very specific sample, a very specific type of person.鈥�
In addition, the study can only reflect the algorithm鈥檚 influence over the six months that the researchers looked at. But Facebook can and does tweak its algorithm regularly. As David Lazer, a political scientist at Northeastern University in Boston, points out in a in Science Express, Facebook announced changes just last month that could skew results 鈥� for example, by programming a preference for updates from 鈥渢he friends you care about鈥�, who may tend to be more ideologically aligned with you.
Facebook should get credit for giving a glimpse of how its algorithm works, says , a computer scientist at the University of Illinois in Urbana. After last year鈥檚 uproar over the company鈥檚 emotional contagion study 鈥� when it manipulated the amount of positive or negative news content to see the impact on posting behaviour 鈥� the company could have easily decided not to publish its research.
Algorithm alterations
Still, many people may be unaware that their Facebook experience is altered by algorithms at all, even by the modest amount that the researchers found, says Karahalios. She suggests that the site makes people more aware of the fact that their clicks and likes have consequences.
鈥淚n a world where we鈥檙e shaped by algorithms, we need to give people insights into the provenance of their information and how it鈥檚 getting there,鈥� she says. 鈥淏y letting them know how it happens, they can make the decision of what they see in their feed.鈥�
It鈥檚 also worth asking why the algorithm screens out conflicting news stories at different rates for liberals and conservatives, says Diakopoulos. 鈥淚f this is a result of how the algorithm is tuned, maybe they need to consider retuning it.鈥�
The study refutes the idea that algorithms alone are creating bubbles around us, says , director of the Center for Civic Media at the Massachusetts Institute of Technology.
鈥淲hat the study further suggests is that my vague hope that Facebook and others might try to engineer diversity and serendipity is unlikely because it would be dreadful for their business model,鈥� Zuckerman says. 鈥淲e display a strong preference for ideologically comfortable content.鈥�
Social networks like Facebook are in a unique position to address that problem, though that doesn鈥檛 seem likely to happen.
Journal reference: Science, DOI: