What is political weight on Twitter really worth?
During the 2022 presidential election, candidates with strong Twitter influence ultimately scored well below their weight on the social network… This is one of the results of Linkage, an innovative social media analysis platform.
If the presidential election is often described as a battle, Twitter has become its most noisy and chaotic battleground. However, many questions arise about the real significance of the social network, or at least its representation. Research is developing tools that allow us to see things more clearly, without taking sides.
Thus, Pierre Latouche, professor at the University of Paris Cité and member of the Laboratory of Applied Mathematics in Paris 5 (Map5), specialist in artificial intelligence and automatic learning, addresses these questions from a mathematical point of view, especially through statistics. His two main research topics are high-dimensional problems, i.e., when the mass of data reaches a point where the complexity of their processing explodes, and‘analysis of networks, especially social networks. participated in its development Connectionsoftware platform managed by four research institutions: CNRS, Inria and the universities of Paris Cité and Côte d‘Azure.
Illustration made from Linkfluence data. They were collected from more than 50,000 Twitter accounts that mentioned the presidential candidate from March 3 to 21, 2022, that is, about 11.6 million messages.
“We started from the observation that people mostly share text on social networks. Pierre Latouche explains. There are tools for analyzing texts and others for exploring networks, but none have combined these two aspects simultaneously. Connectivity is a big step forward. It is indeed one of the first programs dedicated to the analysis of relationships and the circulation of texts in social networks. »
The relationship is built around probabilistic and graphical models and is based on the so-called variational approximation and mean-field methods. Communication is especially placed for learning 2022 presidential electionscooperates with the newspaper The world who drew several articles two made headlines. ” The idea came from Joe Biden’s campaign team’s strategy for the 2020 US presidential election. Pierre Latouche explains. Instead of trying to pressure Donald Trump’s overrepresented supporters on Twitter, the Democratic headquarters decided not to respond to them and only celebrate their successes and the positive elements of Joe Biden’s campaign. Together with the journalists, we wanted to know that it has paid off since he won the election The worldWhat was the real political weight of Twitter in France. »
Every word of every tweet has been scrutinized
To achieve this, Linkage enlisted the help of Linkfluence. It swallowed all tweets that mentioned at least one of the official presidential candidates between March 3 and March 21, 2022. So the company collected about 12 million public messages, which it made available to Linkage for free to feed and shape its models. fascinating network. Internally, each node represents a Twitter account associated with any other nodes it mentions or retweets. Historically, we have had powerful computer tools that can create clusters by analyzing the connections of a network for fifteen years. fronted by Pierre Latouche. Today, using Linkage, we can also analyze the content of exchanges. Thus, every word of every tweet is scrutinized and linked as a whole. »
This work revealed about two dozen groups, called majorities, which the researchers were able to link, in most cases, to political parties participating in presidential elections. Mathieu Jacomy, a researcher at the Tant Laboratory in Copenhagen, then designed and made this data legible. Large map of French political Twitter.
In 2022, the big gap between the twittosphere and reality…
The four central candidates are clearly Eric Zemmour, Emmanuel Macron, Jean-Luc Mélenchon and Valerie Pecresse. However, in the polls, Eric Zemmour and Valérie Pécresse ultimately punched well below their weight on Twitter. The Left was essentially associated with Recalcitrant France, while the Socialist Party and EELV were not at all. On the right, c‘National Rally, which represents 20% of all accounts, is mainly a leading Reconquête party, including above.
An example of a mapping made by Linkage. The colors of the nodes reflect the colors of the found groups and the edges of the automatically estimated discussion topics.
” This is an absolutely huge number.Pierre Latouche kicks. Eric Zemmur’s supporters were probably built around a very effective systemastroturfing, inspired by Donald Trump. It was about creating an activity in which each message was retweeted by Eric Zemmour’s lieutenants, who systematically retweeted each other, before a second circle of followers retweeted intensively in turn. The training effect then creates an extremely important activity. To a lesser extent, Jean-Luc Mélenchon’s entourage used a similar technique.. »
…but in 2017 it was the opposite
Another lesson of connectivity is that the political dynamics on Tweeter have gradually drifted away from the dynamics of society. Maybe because of these tactics. ” In 2017, Twitter visibility and election results were strongly correlated, Pierre Latouche notes. The candidate’s share of supporter activity deviated by only 0.1% from the score obtained in the first round. This year, obviously not at all, the difference between political weight on Twitter and political weight in the real world is important. »
Linkage has thus demonstrated its interest in studying social networks. About 1,400 users have registered on the platform, which currently operates as a research program. ” Link is a live research project, Pierre Latouche concludes. We’ve entered a phase of real expansion, hiring both fundamental research and software development engineers. In particular, we’re working on visualization so that buttons and options make Linkage easier to use, and we’re continuing to add features to it.» ♦
Also read on our website:
The Internet, a way of disinformation?
Social networks: teeth of opinion manipulation
