Giorgia Meloni, leader of the Brothers of Italy party, takes a selfie at the end of an election campaign rally in Caserta, Italy, on 18 September, 2022, ahead of the 25 September election. Credit: Alessia Pierdomenico/Bloomberg via Getty Images.
Political parties are increasingly investing in social media campaigns in an attempt to shape consensus. Refined analysis and complex techniques are now widely used in political communication campaigns1. But the communication between parties and citizens, mediated by social media, is totally asymmetric. Voters usually receive information passively as targets of campaigns that aim to convince them, or to engage them with personalised content.
Researchers and data scientists can help citizens to widen their perspectives and have a less passive role in political communication. In recent years, many studies have probed the complexities of political communication on social media 2 3, but only a tiny part of this knowledge reaches the public. To help close this gap, in the aftermath of the last general election in Italy, we released a report "Elezioni Italiane 2022” thanks to a collaboration between Sony Computer Science Labs of Rome and Paris, the Centro Ricerche Enrico Fermi and Sapienza - Università di Roma. We analysed social media with the aim of offering interactive tools to improve public awareness of political communication. Our results provide a blueprint for applying data science to political communications in future elections and in other countries.
We collated Facebook posts from 2019 to the 2022 elections, and their interaction counts (likes, reactions, tags, etc.) from the public pages of candidates and local newspapers. The latter are often neglected in this kind of analyses, typically more focused on national news outlets. We investigated their role as a link between political forces and the local territories by analysing the similarity of language between the main candidates and regional news. Such similarity was measured by quantifying the alignment between the frequencies of specific words in candidates’ Facebook posts and in those by local news outlets aggregated by region. As a result, a clear difference emerged between locally-oriented candidates (e.g. Luigi De Magistris, former mayor of Naples) and nationally-oriented candidates. Still, even the latter are not uniformly similar across regions, as regional news in some regions presents higher similarity with the majority of candidates. Lombardy, in particular, scored the highest average similarity, pointing out its central role in the political language, while Sardinia scored the lowest.
A more detailed comparison shows language differences between politicians. We use statistical methods to represent the distinctive word of every politician. For example, Giorgia Meloni (the leader of Fratelli d’Italia and current prime minister) and Enrico Letta (then leader of the Democratic Party) were strongly differentiated by the usage of words like ‘immigration’ (frequent for Meloni and unused by Letta) or ‘Ukraine’ (frequent for Letta and unused by Meloni). The same analysis on the aggregated language of all politicians, before and after the start of the electoral campaign, showed the shift in the political debate. Words that were very frequent before the electoral campaign, like “Putin”, became less relevant and were replaced by words like “economics”.
Political language is explicitly designed to trigger a reaction in its target audience, but the actual reaction –or target– doesn’t always coincide with the expected one. We studied how users reacted to the posts published by each candidate, and how the corresponding distribution of reactions changed during the electoral campaign. We observed how some politicians tend to engage their targets on specific kinds of reactions while others play with all the spectrum. We measured the evolution of the situation due to the elections and found that the former Foreign Affairs Minister and former Five Star leader, Luigi Di Maio was the politician with the most important evolution (with a strong increase in ‘angry’ reactions and a strong decrease in ‘love’ reactions), while the most stable was Silvio Berlusconi. This situation might reflect the reaction of the followers to important events of the electoral campaign, like Di Maio’s breakaway from his former party.
After the interaction between politicians and their followers, we studied the interaction of politicians between themselves. The mechanism of tags allows them to address competitors and allies directly, and allows us to build the network of who-tags-who. We found that almost every politician had a strong tendency to tag themselves, and a general link asymmetry. In some particular cases, many links pointed unreturned to a specific politician (e.g. Giuseppe Conte, the former Prime Minister). More in general, we studied the asymmetry between tags sent and received. Giorgia Meloni emerged as the most tagged politician, while Carlo Calenda’s campaign was the one that most often tagged others. The most important asymmetries were scored by the former Minister of Health Roberto Speranza (more than 60 tags received per tag sent) and by Gianluigi Paragone (0 tags received versus 257 sent). The emerging dynamics seem to indicate that tags are used to criticise a focal figure (like a minister) or to try to gain visibility by tagging popular candidates.
The special report ‘Elezioni Italiane 2022’ aims to provide an interactive exploration of data analyses. By helping citizens to understand how politicians focus on specific regions instead of others, which keywords they choose for their narratives, or which reaction they aim to engage in their target, we hope to support them in the decryption of political communication complexities.
This analysis lays the basis for future works on the modelling of political opinion dynamics on social networks, as well as on the evolution of political narratives. How is the political debate about Brexit and Europe changing in the UK? What is the image of countries like Russia or China in the social media discourse of Western countries? What will the local/national dynamics of information in the next US presidential campaign be? Will political messages be tailored differently for different states? Many tools are available to tackle these questions, but they are not accessible for the public. Our platform represents a first step towards improved awareness of political dynamics in our democracies.
