Twitter Algorithm Further Boosts Politically Right-Wing Posts, New Study Finds
The microblogging site Twitter uses an algorithm that increases the reach of politically right-wing sources more than politically left-wing sources, according to a new study. The algorithm is used by Twitter to personalize the content that appears on its users’ personal timelines.
Cambridge University Professor Ferenc Huszar and colleagues used a randomized controlled group of nearly two million daily active Twitter users chosen by the platform to receive content in reverse chronological order without personalization and a processing group representing 4% of all other accounts with custom timelines to determine whether Twitter’s algorithm is amplifying content left or right.
The authors also examined the algorithmic amplification of 6.2 million new policies shared in the United States. According to the results, the personalization algorithm helped amplify tweets from politically right-wing sources in six of the seven countries. Right-wing news services in the United States have also received special attention. Contrary to popular belief, the far left and far right media have not been significantly amplified compared to politically moderate publications. The findings, the authors say, add to the evidence-based discussion about the impact of personalization algorithms in determining consumption of political content.
A similar study, conducted between April 1 and August 15, 2020, covered 5% of the company’s active users and may be one of the largest studies of its kind conducted by one of the leading social media companies on the way their products influence what people read and interact with online, and therefore its societal impact. It comes just a week after a whistleblower identified Facebook’s engagement-based algorithm as a major source of hate speech and fake news. Twitter’s algorithm bolsters right-wing political narratives, the study finds.
“Why these observed models exist is a much more difficult subject”
“… defining why these observed patterns exist is a much more difficult subject,” Rumman Chowdhury, director of software engineering at Twitter, said in a series of tweets describing the results. The Twitter employee added in another tweet: “Twitter is a socio-technical system – our algos are responsive to what’s going on. The next step is a root cause analysis – is this unintentional model bias? Or is it a function of what and how people tweet, and things going on in the world? Or both? “
Here is what we found (part 1): 2 / n
Tweets about political content from elected officials, regardless of party or whether the party is in power, see algorithmic amplification compared to political content on the reverse timeline.
– Rumman Chowdhury (@ruchowdh) October 21, 2021
Experts have urged companies like Twitter and Facebook to share access with university researchers to learn more about the potential problems their products can cause. Concerns have arisen following a series of controversies involving virtually all of the major social media companies, the most notable of which is Facebook.