Is Twitter’s Algorithm Actually Biased in Favor of Right-Leaning Content?

Despite a constant stream of rhetoric asserting otherwise, a new study conducted by Twitter found that right-leaning news outlets see greater algorithmic amplification on its platform compared with left-leaning news outlets, and tweets posted by accounts from the political right receive more algorithmic amplification than the political left when studied as a group. The social network analyzed millions of tweets from April 1 through Aug. 15, 2020, from accounts operated by elected officials in seven countries. Those accounts were defined as follows: Canada: House of Commons membersFrance: French National Assembly membersGermany: German Bundestag membersJapan: House of Representatives membersSpain: Congress of Deputies members.U.K.: House of Commons membersU.S.: Official and personal accounts of House of Representatives and Senate members Twitter said it used this data to test whether or not tweets were amplified more when the timeline is set for algorithmic ranking than when it is set for reverse chronological feed, and whether there was variance by party, and it turned to public, third-party sources, including official institutional websites, to determine party affiliation. As for news outlets, Twitter analyzed hundreds of millions of tweets containing links to articles that were shared during the same time period, categorizing those news outlets based on media bias ratings from independent organizations Ad Fontes Media and AllSides. Twitter said the research was conducted by: Luca Belli (Twitter staff machine learning researcher); Moritz Hardt (University of California-Berkeley, the author was a paid consultant at Twitter, and the work was performed while consulting for Twitter); Ferenc Huszár (Twitter, University of Cambridge); Sofia Ira Ktena (now at DeepMind Technologies); Conor O’Brien (Twitter); and Andrew Schlaikjer (Twitter). The social network also found that tweets about political content from elected officials saw algorithmic amplification compared with political content, regardless of party affiliation or which party is in power, on the reverse chronological timeline. Twitter added that group effects did not translate to individual effects, meaning that two individuals in the same political party would not necessarily see the same amplification. Belli and Twitter director of software engineering Rumman Chowdhury wrote in a blog post introducing the research, “Since 2016, people on Twitter have been able to choose between viewing algorithmically ordered tweets first in the home timeline or viewing the most recent tweets in reverse chronological order. An algorithmic home timeline displays a stream of tweets from accounts you have chosen to follow on Twitter, as well as recommendations of other content we think you might be interested in based on accounts you interact with frequently, tweets you engage with and more. As a result, what an individual sees on their home timeline is a function of how they interact with the algorithmic system, as well as how the system is designed.” They added that the social network asked the following questions while conducting this study: How much algorithmic amplification does political content from elected officials receive in Twitter’s algorithmically ranked home timeline versus in the reverse chronological timeline? Does this amplification vary across political parties or within a political party?Are some types of political groups algorithmically amplified more than others? Are these trends consistent across countries?Are some news outlets amplified more by algorithms than others? Does news media algorithmic amplification favor one side of the political spectrum more than the other? Belli and Chowdhury concluded, “It’s important for us to share the data we used to conduct this study so that other researchers can reproduce our work. To aid this, we are making aggregated datasets available for third party researchers who wish to reproduce our main findings and validate our methodology, upon request. Details on what is included in this data are given in the paper. For full transparency, researchers would ideally have access to the raw data from which these aggregates were calculated, but this is extremely difficult without compromising privacy. We hope that by sharing this analysis today, we can help spark a productive conversation with the broader research community to examine various hypotheses for why we are generally observing comparatively more right-leaning political amplification of elected officials’ content on Twitter.”