Google Scholar's Filter Bubble: An Inflated Actuality?

Google Scholar's Filter Bubble: An Inflated Actuality?

Ke Yu (Human Sciences Research Council, South Africa), Nazeem Mustapha (Human Sciences Research Council, South Africa) and Nadeem Oozeer (SKA South Africa/AIMS/NWU, South Africa)
DOI: 10.4018/978-1-5225-0830-4.ch011


This chapter investigates the allegation that popular online search engine Google applies algorithms to personalise search results therefore yielding different results for the exact same search terms. It specifically examines whether the same alleged filter bubble applies to Google's academic product: Google Scholar. It reports the results from an exploratory experiment of nine keywords carried out for this purpose, varying variables such as disciplines (Natural Science, Social Science and Humanities), geographic locations (north/south), and levels (senior/junior researchers). It also reports a short survey on academic search behaviour. The finding suggests that while Google Scholar, together with Google, has emerged as THE dominant search engine among the participants of this study, the alleged filter bubble is only mildly observable. The Jaccard similarity of search results for all nine keywords is strikingly high, with only one keyword that exhibits a localized bubble at 95% level. This chapter therefore concludes that the filter bubble phenomenon does not warrant concern.
Chapter Preview


This chapter originates from the publication of the 2011 TED talk by Eli Pariser (Beware online ‘filter bubble’1) and some subsequent exchanges among researchers of anecdotal observations that searches on Google do return varied search results. In an attempt to understand this issue, we started investigating the features and determinants of Google’s personalising algorithms. As researchers from the Southern Hemisphere (from now on referred to as South), we asked whether such filter bubbles might impact on our own research processes. We wanted to understand whether the most important personalising determinant—location—meant that our access to knowledge produced elsewhere would be restricted. Moreover, we wanted to understand whether this would influence the quality of our research due to potential variance in our search results. Ultimately, we also wanted to understand whether it would impact the potential success to make our voices heard and compete with scholars from the Northern hemisphere (North) on an equal footing in the global knowledge production.

We decided to investigate these questions by conducting experiments on the extent of the alleged bubbles as well as a survey to understand researchers’ academic search behaviour regarding Google products. We only focussed on the use of Google Scholar for this experiment, as a generic Google search often produces advertisement related information instead of academic information that we are interested. In this chapter, we report on the results from this exploratory experiment, varying variables such as disciplines (Natural Science, Social Science and Humanities), geographic locations (north/south), and levels (senior/junior researchers). We also report on the results from the short survey on academic search behaviour that accompanied the experiment2.

The chapter is organised as follows. The introduction outlines the arrival of the information age and the emergence of Google within this context. We also provide some background on Google and Google Scholar’s usage in the academic environment and the findings on comparing search effectiveness from Google and other academic search engines. We then turn to the main focus of this chapter on Google Scholar’s alleged filter bubbles. We provide a brief review on other studies investigating the extent of Google’s bubbles, the methodology we used for our experiments, our design and method of analysis. Then we report our results, as well as that from the search behaviour survey. We conclude by discussing the limitations of the study and implications of our findings.

Complete Chapter List

Search this Book: