YouTube's recommendation algorithm is left-leaning in the United States

Area of Science:

• Online political communication
• Algorithmic bias in social media
• Computational social science

Background:

• YouTube is a major platform for online political video consumption.
• Its recommendation algorithm significantly influences user exposure to content.
• Concerns exist about algorithmic echo chambers and extremist content delivery.

Purpose of the Study:

• To investigate potential political leanings of the YouTube recommendation algorithm.
• To determine if the algorithm exhibits bias toward the Left or Right.
• To understand the implications of algorithmic political bias on users.

Main Methods:

• Constructed six archetypal user personas across the US political spectrum (Far Left to Far Right).
• Conducted a controlled experiment over eight months, analyzing over 120,000 recommended videos.
• Assessed the algorithm's impact on user consumption patterns and content exposure.

Main Results:

• The algorithm exhibited an asymmetric pull away from political extremes, stronger for the Far Right than the Far Left.
• YouTube's recommendations showed a leftward skew, even for users without prior watch history.
• Algorithmic bias was observed, influencing content exposure across different political leanings.

Conclusions:

• YouTube's recommendation algorithm demonstrates a political bias, skewing leftward.
• This bias is asymmetric, impacting users differently based on their political orientation.
• The findings raise critical questions about the societal and political implications of biased algorithms on platforms like YouTube.