Listen to this story
Twitter acknowledges that its algorithm favors right-wing tweets
The company did a study on the amplification of political content on the platform
Twitter is one of the most controversial platforms, many say that on Facebook you make friends and on Twitter you fight with them. This reputation of being a social network — boxing ring, has been gained in good part by so many discussions about politics.
Either because of this reality or as a study to analyze the results of its algorithm on the political issue, Twitter carried out a fairly complete analysis whose results have impacted many: The algorithm widely favors the right on Twitter.
Seven countries selected for the study
The selection is wide and involves seven countries from 3 continents: Japan, the United Kingdom, France, the United States, Spain, Germany and Canada, these are the countries object of the analysis and specifically, media with a clear political tendency, as well as political representatives of these sectors.
The platform specialists, taking into account that the political representatives do not have a constant presence on the web, decided to add the interaction and projection of the media content, selecting media that are clearly identified with some trend.
Luca Belli, Andrew Schlaikjer, Sofia Ira Ktena, Ferenc Huszár, are part of the Twitter team that designed and carried out this interesting study.
At the political level, tweets / content of: members of the House of Commons in Canada, members of the French national assembly, accounts of members of the United States House of Representatives and Senate, representatives of the German Bundestag, members of the House of Representatives of Japan, representatives of the House of Commons of the United Kingdom and members of the National Congress of Spain.
What questions were raised for the study?
Although a study of this type is much more complex than thinking about some questions as a basis, it necessarily starts with questions that define the direction to take. In this case, Twitter investigated:
How much algorithmic amplification does political content receive from elected officials?
Does this amplification vary between political parties or members of the same party?
Are some political groups algorithmically amplified more than others within the same country?
Are some media more than others amplified by the algorithm due to their political bias?
Are these trends constant in various countries?
And of course, the question that rounds up all the above: Does algorithmic amplification on Twitter favor one political trend more than another?
Timeline amplification vs chronological amplification on Twitter
All the previous questions are made by contrasting the result in both configurations.
Since 2016, the blue bird network offers the possibility that we determine how we want to see the content in our timeline, the options are exactly these: A timeline that from the beginning shows us results for the people and content that we follow, or a reverse timeline (which has the advantage of helping you not miss tweets that interest you).
The Twitter study started in 2020
The study time is from April 1 to August 15, 2020. The reason that until now we have access to the complete data of this study, are due to the need to order data and analyze all the material collected during those months.
No one is fooled by the chronological duration of the study, in that time hundreds of millions of tweets were analyzed in the seven countries mentioned at the beginning. Filters were used to not process tweets with reference to topics such as sports or cooking.
Results of the Twitter study on the algorithm and political trends
The results not only indicated that the algorithm favors the right, there is other data that will surely be valuable for future changes or adaptations on Twitter. Specifically, the results of the study indicate that:
Political content is further amplified in timeline settings based on the interests of the user in a direct way, compared to the timeline in reverse chronological sense, which rescues and shows us content in a less select way.
Personification is imposed on the results of the algorithm. Twitter offers a personalized experience for specific user tastes.
For this reason, some group results in amplification of parties or political representatives are not reflected in the same way in each person who follows or identifies with that political group.
In six of the seven countries, tweets linked to the political right receive more amplification than those to the left, the exception to this case is Germany, the only country where the right does not have that "advantage given by the algorithm."
The right-wing media also benefit in all the countries studied by the Twitter algorithm, in relation to left-wing political content (At this point Twitter saves its responsibility regarding the classification of the right-left media, some of what third organizations deal with).
Twitter actions from the study of its algorithm and the imbalance it presents
In the first place, they make it clear that they cannot specifically determine the reason for this imbalance in the algorithm, but they undertake to review the situation based on a clear principle: algorithmic amplification is not the problem since all algorithms amplify. The problem is when this amplification widely benefits one sector over another, and that is precisely what is happening on the platform.
This study was carried out with the direct aim of knowing the amplification and if it benefited one political sector over another.
Now that these results are handled, it is to be imagined that on Twitter they will do other studies to find out the causes of this inclination in favor of a political tendency.
It is a valuable work that pursues something like the impartiality in the behavior of the algorithm in a social network. Something very complex, but worth the studies you need.
The privacy that a social network must guarantee is an obstacle, but respecting these limitations, Twitter will work on the basis of guaranteeing several rights of its users: privacy, responsibility of the platform and impartiality in the treatment and amplification of content on the network.