Twitter’s algorithm favors the right and doesn’t know why

The algorithm that determines the Twitter timeline is a right-wing algorithm. Or it seems that way, without the company itself knowing why it is not politically neutral. An internal investigation of this social network with elected politicians in seven countries (Germany, Canada, Spain – members of Congress – and the United States, France, Japan and the United Kingdom) concluded that in six of them – with the exception of Germany – “receive tweets posted by accounts on the political right more arithmetic inflation than those on the political left when studied as a group.” The People’s Party is one of the groups that has benefited most from this anomalous process.

In the part of the study that corresponds to Spain, it was noted how PP had one of the highest inflations


The social network also explained that right-wing media “receives greater computational amplification on Twitter than left-wing media,” said its director of software engineering, Roman Choudhury, and researcher on the official blog. in machine learning. Luca Bailey.

Twitter admits that determining why the right patterns occur is “a lot harder to answer”

The company discloses that “certain political content is amplified on the platform,” but notes that “determining why these observed patterns occur is a difficult question to answer, as it is the product of interactions between people and the platform.” The mission of the internal team called META (Machine Learning Ethics, Transparency, and Accountability), according to Twitter, is to “identify and mitigate both disparities that may occur.”

See also  Lenin Castillo and his terrible knockout against Callum Smith | Video | Dominicans recover | Sports

For the report, Twitter analyzed the “millions of tweets” of the accounts of these elected politicians between April 1 and August 15, 2020 to see if each tag’s messages were further amplified in the social network’s timeline, which was categorized by the algorithm. To determine the media amplification of each label, it was ranked by the independent organizations AllSides and Ad Fontes Media. Tweets that referred to non-political content, such as recipes or sports, were excluded.

The company will continue to study the phenomenon with outside researchers to try to find out why.

The Twitter research team made the content available to outside big data researchers to validate their findings. The company wants to take advantage of this dialogue with other experts to “examine various hypotheses as to why we generally observe a relatively more right-wing political amplification” of the content of elected politicians on Twitter.

Leave a Reply

Your email address will not be published. Required fields are marked *