Split Screen cannot reverse-engineer Facebook’s recommendation algorithm, and none of our observations should be treated as causal claims that Facebook has targeted a specific piece of content at a specific group. Using data collected from our Citizen Browser panel, we are only able to demonstrate what Facebook’s recommendations algorithm surfaced to users aggregated by their self-disclosed political leanings, gender, and age. For more information on the limitations, see our methodology.

Content shown on Split Screen is not necessarily the most seen or interacted with by our panelists. We compare the percentage of each group that was served each piece of content. The bigger the difference in percentage between each demographic in a pair, the higher the rank, or position, of the content.

If you see something that you didn't expect and think we should know about it, email us at splitscreen@themarkup.org

• Concept and development Sam Morris and Surya Mattu
• Editor Julia Angwin
• API and Data Pipeline Developer Micha Gorelick
• Data Processing Jon Keegan
• Technical Coordinator Angie Waller
• Data Pipeline Developer Leon Yin
• Header and Avatar Illustrations Bratislav Milenković
• Infrastructure and Security Engineer Simon Fondrie-Teitler
• Copy Editor/Producer Jill Jaroff
• Special Thanks Ian Ardouin-Fumat, Jeff Crouse, Mago Torres, Netograph, Ronald Robertson