The Limited Effects of the Personalisation of Search

The second day at ICA 2018 starts for me with a panel on the personalisation of search, and the first presenter is Grant Blank. He begins by noting the importance of free-flowing information for society, but of course the media through which such information flows have changed over time, and this has affected media biases. Contemporary media now form a diverse media ecology.

Do online media in their diversity empower citizens to make better-informed decisions, then, or does the personalisation of online media distort the information that citizens encounter? Much of the present discussion is severely undertheorised. There is a technologically determinist discussion of echo chambers and filter bubbles; and a moral panic about how technologies promote ‘fake news’. There are also socially determinist perspectives on echo chambers, confirmation bias, winner-takes-all effects, and the spiral of silence. And there are mixed perspectives that examine agenda-setting processes and the two-step flow of information from more to less informed people.

The present project (and indeed the entire panel) uses data from the Quello Search Project, drawing on a random sample of 14,000 Internet users across seven nations in January 2017; these completed an online survey about their search practices. Search was one of the first major challenges of the early Internet, with Google now clearly established as the leading solution for finding information online; 86% of Internet users use search daily.

The average person with an interest in politics looks at 4.5 sources per day, roughly split across online and offline sources, and search is the major pathway to such sources. 36% of users read news they disagree with either often or very often, and 15% say their friends have different and 65% say their friends have mixed beliefs. Users also check and confirm information frequently, and they use search actively, especially also to check information encountered via social media. 75% of users in Italy say they check facts daily, for instance. This can also change people’s opinions. People rarely block or unfriend others over political disagreements.

The data do not support deterministic perspectives, then. There is a social shaping of Internet use, but not to the complete exclusion of other views; what emerges as central here are search skills and overall digital media literacy. Any risks of personalisation that do emerge may be able to be addressed via behavioural nudging of users.