Access to high-quality information is a central tenet of democratic regimes, fundamental to help citizens make educated decisions; Conversely, false information can be used to manipulate and distort democratic processes.

The internet is increasingly a top source of information, but how this information is ranked is not clear: social networks and search engines use mostly proprietary algorithms to promote posts of pages according to the users’ preferences. Since content is delivered based on what is deemed more ‘interesting’ or ‘relevant’ for a user, there is no guarantee it is the most accurate.

Through this project, we will focus on how online search-engines use information collected outside of their platforms (for example while browsing online) and develop a tool to audit them.

Specifically, we use a system of bots, or webcrawlers, to generate different browsing histories and test how the algorithms of different search engines use these histories to rank their suggested results.

In particular, we aim at answering questions such as:

1. How do browsing history influences online profiles?
2. How do these profiles influence how search-engines rank their results?
3. How common is it for search-engines to expose users to disinformation?

Importantly, as our system does not rely on real user data, it raises very few ethical concerns and can be rapidly adapted and scaled to cover many countries and topics.

Once developed, our auditing will allow NGO’s, journalists, interested users and even the search-engine platform themselves to recognize recommendations for low-credibility content pages.

The findings will shed light on how browsing history influences search engine results, and the likelihood of being directed to disinformation.

This project is funded by an ERC Proof of Concept Grant.