Written for smalltalkeconomics.com.
Recently, social media have become the primary source of information for many and thus play an increasingly important role in influencing opinions, and apparently more so for the young generation. The problem is that on social media, opinions you usually get exposed to become less and less diverse as a result of both automatic filtering algorithms (whose goal is to get as many shares as possible) and the fact that you are more likely to follow people with similar views (who are, in turn, likely to share information biased in favor of their opinion). An example is your Facebook feed, which is personalized based on your past clicks and likes, making you less likely to “see the bigger picture”. This phenomenon is called a filter bubble and it concerns both search engine results and social media. However, as some people argue (here or here), social media are a more powerful tool for filter bubbles, and this is a problem. Others think that when it comes to the really important decisions, it might not be that bad just yet, and there are even guidelines on how to get out of your filter bubble.
In a recent working paper, Shane Greenstein, Yuan Gu and Feng Zhu (all from Harvard Business School) analyzed whether Wikipedia, and in particular articles about US politics, are also affected by filter bubbles and thus become more biased over time. They chose politics because it is a good example of the so-called contested knowledge, which can be loosely defined as knowledge that answers questions for which there is no single “right answer”. Relying on metrics developed by Gentzkow and Shapiro’s 2010 article in Econometrica, the authors measure empirically whether selected Wikipedia articles become more or less segregated (i.e. slanted towards a certain view) over time.
Their results, somewhat optimistically, show that Wikipedia contributors increasingly offer content to those with different points of view, which reinforces unsegregated conversations at Wikipedia over time. Interestingly enough, the authors additionally estimate that this slant convergence process takes one year longer on average for Republicans than for Democrats. The authors stress the importance of the option to remove previously added material (such as on Wiki-style pages) or using aggregate contributions (such as Yelp or Rotten Tomatoes) over the social media style, where additional material just gets piled up on top of what is already there.
Collective intelligence, of which Wikipedia is perhaps the most astounding source, thus seems to cope fairly well with the filter bubble problem, at least for now. While we can, let’s all try to use that to our advantage and rely more on objective sources when forming opinions, challenge ideas shared by people we follow on social media, verify facts using multiple sources and most importantly, keep an open mind. And share this article with our filter bubbles.
Reference: Greenstein, S., Gu, Y., & Zhu, F. (2016). Ideological Segregation among Online Collaborators: Evidence from Wikipedians (No. w22744). National Bureau of Economic Research. Available here.