The UK General Election result in May surprised me; I did not expect a Conservative majority, nor did I expect UKIP to get so many votes. But why was I surprised?

Being in Mozambique, most of my UK news comes through the few news websites I visit (BBC News and The Guardian), Twitter and Facebook. In the lead up to the election, nearly all of these sources told me to expect a multi-party coalition, a form of government that we haven’t had for many years, and perhaps a farewell to the usefulness of first-past-the-post in the 21st century. I became convinced; the more I read, the more I expected this result. From the fallout and vocal complaints made by many left-leaning people, it seems like they too expected this result, and the Conservative majority shocked them into protest. The Guardian lead with headlines like “David Cameron wins surprise majority in general election”.

I asked myself, “How could this be?” Disbelief that my views are only shared by 30% of the population, that the rest didn’t respond to the social justice narrative I had consumed in my daily online news updates. A friend shares her similar experience on the election night, and my Facebook was filled with much of the same.

A new paper in Science sheds light on these ‘surprise’ results, using US data from Facebook (links shared and ideological affiliation) to explain the phenomena that has lead to people like me being so unprepared for the results of a national election.

The data in the paper is taken from Facebook’s News Feed – the authors themselves have Facebook affiliations (which I thought would count as a conflict of interest). The News Feed is not a simple timeline of every post from your friends. Rather, it generates a selection of posts algorithmically, based on what it thinks you will find most engaging. This feedback results in the tendency to create an “echo chamber” – showing you only those posts which reflect your own attitudes and opinions. It means that we tend not to see viewpoints that disagree with our own perspective; we don’t see cross-cutting content.

The authors wanted to find out how much the Facebook News Feed algorithm is involved in this echo chamber effect. The results show, however, that the influence of the algorithm is slight; viewpoints from the opposite side show up just 1 percent less often than if the News Feed was not justing the figures. The real echo chamber effect comes from ourselves: people click on challenging news less often than they see it.

Figure 3B shows the percentage of cross-cutting content seen by ideologically conservative or liberal people, i.e. the percentage of conservative content seen by liberals and the percentage of liberal content seen by conservatives. When all the posts shared on the News Feed are taken into account, liberals have the potential to see more cross-cutting content than conservatives, reflecting the greater number of conservative posts. However, liberals filter this potential more strongly than conservatives, as shown by the steeper drop in percentage. Liberal feeds (the potential from the network) contain about 24 % cross-cutting content, whereas conservative feeds contain about 35%, indicating that liberals are connected to fewer cross-cutting friends than conservatives. The News Feed filter exposes them to almost all of that content (23% and 34% respectively), but further filtration occurs at selection – the number of pieces of content that is actually clicked on. Liberals end up clicking on only 20% of the cross-cutting content; conservatives on 29%. It’s a defeat for News Feed haters.

Though this research studies American political views from American citizens, I imagine the results would be similar in the UK. Is it any wonder then that I was surprised by the election results? I have built around myself a chamber of echos, a positive feedback loop reinforcing my own views. The “shy tories” may not have been that shy at all – I was just not connected to them in the first place, and those that I was, I ignored.

Featured research: 

Bakshy et al., 2015, Exposure to ideologically diverse news and opinion on Facebook. Science.