Project Description

  Search engines, such as Google, are critical information intermediaries in today's digital media ecosystem. By prioritising specific information sources and content items in response to user queries, search engines structure social reality by determining how societally relevant issues, such as popular votes, are presented to the public. However, the performance of search engines can be biased and result in a systematically skewed representation of different issues, for instance, due to search engines prioritising a small selection of sources or giving disproportionate visibility to a single viewpoint. Such a skewness can undermine citizen rights, particularly when bias results in the promotion of misleading information and amplifies the fragmentation of the public sphere by exposing users from different language communities to drastically different interpretations of the same issue.

  The project "Algorithm audit of the impact of user- and system-side factors on web search bias in the context of federal popular votes in Switzerland", led by Dr Mykola Makhortykh, aims to make three contributions to the research on search bias in the context of political communication in Switzerland. First, the project introduces a more nuanced conceptualisation of bias to address the difficulties of defining systematic skewness under often absent baselines and to consider whether certain forms of bias can benefit the public sphere. Second, the project examines the impact of user- (e.g. the language of the query and its semantic composition) and system-side factors (e.g. randomisation of search outputs, the impact of earlier browsing history, and time-based changes in source relevance on search bias. Third, the project traces how general and vote-specific user characteristics can affect individual exposure to biased search outputs.

  To make these contributions, the project examines how the two most used search engines in Switzerland - Google and Bing - retrieve information about three rounds of federal popular votes in 2023-2024 in response to search queries in German, French, and Italian. The project uses an innovative methodological design combining survey research, mixed-method algorithm audits, and automated content analysis. It employs surveys to examine the relationship between individual characteristics and user-side search factors and identify what sources and interpretations Swiss citizens expect search engines to prioritise regarding popular votes. The algorithm audits are used to collect search outputs around each vote and to examine the impact of user- and system-side factors on how search engines prioritise information sources. The automated content analysis is used to detect news frames in search outputs to identify whether certain combinations of user- and system-side factors lead to search engines prioritising different sets of frames in relation to individual popular votes.