Political discussions are a constant on social networks, regardless of the country or the names of the parties involved. Anyone who enters Twitter or Facebook you can attest from day to day that political-driven, mostly high-voltage, user-to-user exchanges are common. However, there are also counterpoints on whether the platforms themselves are responsible for promoting content related to a particular political position.
In the last hours, Twitter released the result of a study dedicated to examining the “algorithmic amplification” of political content on the platform. And to the surprise of many (or not), the social network itself confirmed that its algorithms promote content from political figures and right-wing news media.
The investigation took place in seven countries: Spain, France, Germany, United Kingdom, Japan, Canada and United States. In it, the tweets from the accounts of elected officials were analyzed, along with millions of publications from the rest of the users that included links to news media articles. The studied tweets covered the period between April 1 and August 15, 2020, as explained by Twitter.
Another interesting fact to highlight before moving forward with the results of the investigation is that Twitter has relied on third-party sources (public records and institutional websites, for example) to determine the political affiliation of officials. This means that has not taken into account the content of the tweets to determine the degree of range or amplification they receive.
While to categorize the news media, it also made use of outside contributions. In this case they were AllSides and Ad Fontes Media the institutions that were in charge of providing the parameters of political identification in question. In addition, Twitter only analyzed the tweets with links to articles on politics in the media studied, and left out those that included other types of information (sports, for example).
Twitter algorithms amplify right-wing political content
Among the conclusions obtained by Twitter there are two that stand out. First of in six of the seven countries in which the analysis was carried out, the algorithms of the social network give a greater amplification of tweets published by officials on the right than those on the left, when studying them as groups. Regarding the news media, meanwhile, it is also indicated that the algorithmic amplification is greater for those on the right than for those on the left. The only country where results different from the rest were observed was Germany.
What motivates this? Twitter doesn’t have an answer for it yet. Rumman Chowdhury, director of the software engineering area of the social network, published some insights on the subject in her profile on the platform.
“The team did a phenomenal job identifying what is happening. Establishing why these observed patterns occur is a significantly more difficult question to answer. […] Twitter is a sociotechnical system: our algorithms respond to what is happening. What follows is a root cause analysis: is this an inadvertent bias of the model? Or is this a function of what and how people tweet, and the things that are happening in the world? Or both? “He said.
Twitter has been working for a long time to detect and correct biases in its algorithms. But it is still striking that a platform can’t figure out why their own technology behaves in this or that way, at least in certain situations. Chowdhury mentions that since there is no “master algorithm”, but rather an algorithmic system, they must not only detect the existence of the alleged biases, but also their origin and the possible responses to them. “When algorithms are released into the world, we cannot model for what happens when people interact with them,” he said in an interview with Protocol.