Measuring Implicit Personalisation on Google News
Personalisation algorithms have been implemented in the top 3 visited websites, namely Facebook Newsfeed, Google Search and YouTube related videos. In addition to these websites there is a variety of personalised news websites, such as The Washington Post’s Trove.com, The New York Times News.me and Google news.
This recent increase of personalisation worries critics of ‘The Filter Bubble Effect’ or in other words “a hyper personalised fragmented web”. These concerns are strengthened by a lack of transparency in the personalisation algorithms and the unknown way personalization affects the content. While there has been extensive speculation on the implications of a hyper personalised web there currently is little quantitative research on the amount of web personalisation. In this research we measure the amount of implicit personalisation of Google News. Google News is chosen because it claims to connect one billion unique users a week to news content and it has been a personalised news service since its launch. We use the method developed by Hannak et al, for measuring personalisation and contribute to a better understanding of the personalised web. In our research we found differences in the articles on the front page. But only small deviations were caused by personalisation.
Tom Groen, "Measuring Implicit Personalisation on Google News", Master's Thesis for the Media Technology programme, Leiden University (The Netherlands), 2014