Digital democracy European Union

Briefing Paper 117: How Do Social Media Algorithms Rank Content And What Can the Digital Services Act Do About It?

Social media platforms have become public squares, spaces where much of our political conversations occur. Anonymity and zero-cost publishing make these platforms accessible to almost anyone. But this also brings about challenges. Malevolent actors use the platforms to seamlessly spread fake news, hate speech, and extremist ideas. This has harmed political discourses in many countries, contributing to a series of recent political shifts that have put our democracies at risk. Social media hinders our capacity to share a common understanding of basic facts, a prerequisite of any democratic society.

Our briefing paper 'How Do Social Media Algorithms Rank Content And What Can the Digital Services Act Do About It' examines strategies for creating a reliable and safe online speech that guarantees fundamental rights. While previous debates around online speech focused on removing content, attention is now shifting toward a new domain: amplification or ranking systems. These AI-based algorithms determine the order in which a user will see online content. Acting on ranking systems can alter what we see on social media but does not remove content.

Authored by Democracy Reporting International's Executive Director, Michael Meyer-Resende, this briefing paper examines the complex algorithmic mechanisms behind content ranking and ways in which policy could step in to guarantee that social media companies have the public interest in mind when ranking content. Specifically, the EU's Digital Services Act (DSA) will compel social media platforms to identify risks to privacy and the rights of children and establish transparency measures and audits.

Access the brief below to read more on how ranking algorithms work and ways in which the DSA could contribute to content ranking systems that guarantee our fundamental rights.


Briefing Paper 117 - How Do Social Media Algorithms Rank Content Download

This work is supported by

Stiftung Mercator