Algorithmic amplification of politics on Twitter.

Area of Science:

• Social Media Research
• Computational Social Science
• Political Communication

Background:

• Personalization algorithms on social media platforms like Twitter shape content visibility.
• Debate exists on whether these algorithms disproportionately amplify certain political groups.
• Understanding algorithmic bias is crucial for informed public discourse.

Purpose of the Study:

• To quantitatively assess the impact of Twitter's personalization algorithms on political content amplification.
• To investigate whether algorithms favor specific political ideologies or parties.
• To provide empirical evidence on algorithmic bias in political content distribution.

Main Methods:

• Conducted a large-scale, randomized experiment involving nearly 2 million daily active Twitter accounts.
• Implemented a control group exposed to a reverse-chronological feed, free of algorithmic personalization.
• Analyzed tweets from elected legislators across seven countries and the US media landscape.

Main Results:

• In six out of seven countries, the political right received greater algorithmic amplification than the political left.
• Algorithmic amplification in the US media landscape favored right-leaning news sources.
• No evidence was found that algorithms disproportionately amplify extreme political groups over moderate ones.

Conclusions:

• Twitter's personalization algorithms demonstrate a bias favoring right-leaning political content.
• The findings challenge the public perception of algorithms amplifying fringe political groups.
• This research contributes empirical data to the ongoing debate on algorithmic influence in politics.