By Caroline Pena, PhD student; Padraig MacCarron, Assistant Professor and David O’Sullivan, Assistant Professor. MACSI (Mathematics Applications Consortium for Science and Industry), Department of Mathematics & Statistics, University of Limerick, Ireland
We analyse social networks to understand how they affect social interactions, polarisation of ideas and the spread of information. Understanding the dynamics of communication through social media is of great importance as social media has been playing a huge role in shaping people’s opinions and their understanding of what different sides of a debate stand for.
Our new research has revealed that the social media platform Twitter (now known as X) has created polarised environments where the discussion of controversial topics is made even more challenging.
Our paper, just published in the journal Royal Society Open Science, analysed information spread and polarisation on the social network Twitter (prior to it being rebranded as X), with specific case studies on the 2018 abortion and 2015 same-sex marriage referendums in Ireland. It shows a clear distance between users backing the ‘yes’ vote and the ‘no’ vote in each referendum.
We used Twitter data as, at that time, it was a popular platform for the expression of opinion and dissemination of information to identify opposing sides of a debate and, importantly, to observe how information spreads between these groups in our current polarised climate.
To achieve this, we collected over 688,000 tweets from the Irish Abortion Referendum of 2018 to build a conversation network from users’ mentions with sentiment-based homophily (how connected users tended to use similar language in terms of positivity and negativity).
From this network, we were able to use community detection methods to isolate yes- or no-aligned supporters with a high degree of accuracy (90.9%). This was supplemented by the identification and tracking of information spread (which we call information cascades) over 31,000 retweets. Using these cascades, we were able to track not only how information spread across these networks; but who was retweeting whom in terms of their alignment with the yes- or the no- side of the debate (Figures 1 and 2).
Figure 1 Example of what a cascade of information spreading through a network could look like.
Figure 2 The resultant cascade tree structured tagged as yes- and no-aligned voters (blue is yes-aligned and red is no-aligned)
It was notable that we found little information spread between the identified polarised yes- and no- aligned communities (see figure 3). Indeed, information tends to spread heavily inside the same ideological community and less frequently between communities. This means that users tend to communicate mainly with those with whom they share an ideology and less with people who have opposing opinions – particularly regarding the abortion debate in Ireland.
Figure 3 (a) Proportion of Yes users by cascade tree (of the cascades that have 10 or more classified users). Note that the y-axis is on a log-scale. (b) Behaviour of information diffusion by seed community (of the cascades that have 10 or more classified users). (c) Density of Yes users by cascade tree, coloured by cascade seed type (of all cascades).
This research provides a valuable methodology for extracting and studying information diffusion on large networks that contain polarised groups and exploring the propagation of information within and between these groups. Importantly, it shows that a tweet that originated in the yes-voting group rarely spreads to the group voting no, and vice-versa, which makes it challenging for people with opposite ideas to hear each other.
The study, ‘Finding polarized communities and tracking information diffusion on Twitter: a network approach on the Irish Abortion Referendum’, by Caroline B. Pena, Pádraig MacCarron and David J. P. O’Sullivan, has just been published in the journal Royal Society Open Science. https://doi.org/10.1098/rsos.240454