24 March 2017 327 words, 2 min. read

Recommendation algorithms : the myth of filter bubbles at stake

By Pierre-Nicolas Schwab PhD in marketing, director of IntoTheMinds
We’ve already discussed filter bubbles on this blog, a rather technical term that describes the negative effet of algorithms on our digital life. For more information read this thorough article. Whether or not filter bubbles really exist is still debated […]

We’ve already discussed filter bubbles on this blog, a rather technical term that describes the negative effet of algorithms on our digital life. For more information read this thorough article. Whether or not filter bubbles really exist is still debated and I wanted to compile all scientific results on the subject. There are actually very few of them/

Although Pariser (2011) has claimed the obvious existence of filter bubbles, contradictory results have been obtained. Interestingly Anderson (cited in Fleder et al. 2008) wrote :

“The main effect of filters, [which include online recommender systems], is to help people move from the world they know (‘hits’) to the world they don’t (‘niches’)” (2006, p. 109). This is the exact opposite of the effect hypothesized by Pariser (2011).

Bakshy, Messing and Adamic (2015) conducted a research on a sample of 10.1m Facebook users and concluded that the recommendation algorithm implemented on Facebook (Newsfeed algorithm) didn’t lead to polarization of users’ political opinions.

Flaxman, Goel and Rao (2016) conclude on limited effects in terms of segragation on political news consumption (“though the predicted filter bubble and echo chamber mechanisms do appear to increase online segregation, their overall effects at this time are somewhat limited”)

Zuiderveen Borgesius et al. (2015) conclude that there is no risk of filter bubbles, because the technology is still in its infancy. The advantages of information abundance largely counterbalance the drawbacks of personnalisation (which is made necessary by abundance).

Nguyen et al. (2014) studied the influence of a recommender system on individual choices of movies and concluded that recommendations actually increased the diversity of movies watched. As the authors put it, “taking recommendations lessened the risk of a filter bubble” [italics in original].

Conclusion

Despite the surge in the interest of medias around filter bubble, scientific evidences accumulated on the subject tends to prove that filter bubbles don’t exist. The negative effects of algorithms, trapping consumers in past consumption, is offset by the abundance of online content that triggers users’ curiosity.



Posted in big data, Innovation, Research.

Post your opinion

Your email address will not be published. Required fields are marked *