Networked partisanship and framing: A socio-semantic network analysis of the Italian debate on migration

Abstract

The huge amount of data made available by the massive usage of social media has opened up the unprecedented possibility to carry out a data-driven study of political processes. While particular attention has been paid to phenomena like elite and mass polarization during online debates and echo-chambers formation, the interplay between online partisanship and framing practices, jointly sustaining adversarial dynamics, still remains overlooked. With the present paper, we carry out a socio-semantic analysis of the debate about migration policies observed on the Italian Twittersphere, across the period May-November 2019. As regards the social analysis, our methodology allows us to extract relevant information about the political orientation of the communities of users-hereby called partisan communities-without resorting upon any external information. Remarkably, our community detection technique is sensitive enough to clearly highlight the dynamics characterizing the relationship among different political forces. As regards the semantic analysis, our networks of hashtags display a mesoscale structure organized in a core-periphery fashion, across the entire observation period. Taken altogether, our results point at different, yet overlapping, trajectories of conflict played out using migration issues as a backdrop. A first line opposes communities discussing substantively of migration to communities approaching this issue just to fuel hostility against political opponents; within the second line, a mechanism of distancing between partisan communities reflects shifting political alliances within the governmental coalition. Ultimately, our results contribute to shed light on the complexity of the Italian political context characterized by multiple poles of partisan alignment.

Fig 1. Monopartite projection on the layer of verified users.

The network is obtained from the tweeting and retweeting activity of verified and non-verified users across the entire observation period (May-November 2019). The five final communities pivot around verified accounts of the main Italian political parties/coalitions and politicians (i.e. far-right parties as Brothers of Italy and the League, in green; center-left parties as the Democratic Party and Italy Alive, in red; the Five Stars Movement, in yellow; Go Italy in blue) as well as around accounts of media, intergovernmental and non-governmental organizations.

Fig 2. Evolution of the number of tweets and users within partisan communities across the entire observation period.

In general, the trend of tweets (on the left) is characterized by weekly oscillations where peaks, coinciding with relevant political or issue-related events (e.g. the 2019 European elections, the ‘Sea-Watch 3’ crisis, the Italian government crisis) appear. The community producing the highest number of tweets is the DX one, followed by the CSX and the M5S ones. In correspondence of July 2019, a peak characterizing the trend of the number of users (on the right) within each partisan community is clearly visible due to the ‘Sea-Watch 3’ controversy, which also induces a single community of ‘governmental supporters’.

Fig 3. Evolution of the partisan communities at a monthly time scale from July to October.

In October 2019, politicians of the main center-left party (united in the red cluster in September) split into two sub-communities, the orange one being induced by the Twitter activity of the members of a new political formation (i.e. the Italy Alive party).

Fig 4. K-core decomposition of the July 2019 semantic networks for the DX (top left) and the CSX (bottom left) partisan communities.

The k-core decomposition reveals the bulk of the discussion about immigration developed by these two partisan communities: while the innermost core of the DX-induced semantic network (top right figure) is composed by hashtags like #salvininonmollare (‘Salvini don’t give up’), #arrestatecarolarackete (‘arrest Carola Rackete’), #iostoconsalvini (‘I stand with Salvini’), that of the CSX-induced semantic network (bottom right figure) is composed by hashtags like #salvinidimettiti (‘Salvini resign’), #fateliscendere (‘let them get off’), #carolaracketelibera (‘free Carola Rackete’).

Fig 5. K-core decomposition of the July 2019 semantic networks for the M5S (top left) and the MINGOs (bottom left) partisan communities.

The k-core decomposition reveals the bulk of the discussion about immigration developed by these two partisan communities: while the M5S community displays a mixed behavior towards immigration policies, as shown by the hashtags #freecarola and #bibbiano, the MINGOs innermost k-shell uncovers a strong support towards pro-migration positions, as proven by the hashtags #ioaccolgo (‘I host’), #iostoconcarola (‘I stand with Carola’) and #facciamorete (‘let us act as a network’).