It’s a good day to be a mainstream politically conservative party or a right-leaning news outlet on Twitter. The company has released the results of a study analyzing algorithmic amplification of political content on the platform, which affirms what had already been suspected by some: The political right does really well on Twitter.
Twitter’s study, led by the company’s Machine Learning Ethics, Transparency and Accountability team, or META, reviewed millions of tweets of elected officials in seven countries— Canada, France, Germany, Japan, Spain, the UK, and the U.S.—as well as hundreds of millions of tweets containing links to articles from news outlets. In all countries except Germany, the company found that tweets posted by the political right were amplified more than those posted by the political left.
When it comes to news outlets, the same thing occurred. (The company analyzed the links to content from news outlets, not tweets by the news outlets themselves). Right-leaning news outlets received more algorithmic amplification than left-leaning news outlets. Twitter didn’t classify news outlets as left-leaning or right-leaning according to its own criteria, but rather used a classification from third-party researchers.
The study determined that certain political content is amplified on Twitter. In the end though, one of the most important questions remained unanswered: Why?
Rumman Chowdhury, director of the META team, told Protocol on Thursday that some of the amplification could be user-driven, related to peoples’ actions on the platform.
“When algorithms get put out into the world, what happens when people interact with it, we can’t model for that. We can’t model for how individuals or groups of people will use Twitter, what will happen in the world in a way that will impact how people use Twitter,” she told the outlet.
In a Twitter blog post, Chowdhury and machine learning researcher Luca Belli wrote that the META team aimed to examine these issues and mitigate any inequity they may be causing. They added that algorithmic amplification is not “problematic by default.”
“Algorithmic amplification is problematic if there is preferential treatment as a function of how the algorithm is constructed versus the interactions people have with it,” they wrote. “Further root cause analysis is required in order to determine what, if any, changes are required to reduce adverse impacts by our Home timeline algorithm.”