Mapping online political talks through network analysis: a case study of the website of Italy's Five Star Movement

ABSTRACT

Understanding relations among online users involved in political discussions can help us understand similarities and differences with corresponding offline interactions. Online communities generally demonstrate too high a level of homophily among users to be illustrative. But online political discussions do not necessarily prevent diversity of opinions. The discussion Forum of Italy's Five Star Movement provides an interesting case study because of the diverse political orientations of participants and the wide range of discussion topics. I apply network analysis to map in detail the relations among the Forum's users to unravel network characteristics and actor behaviours. After harvesting 86,943 discussions and 461,297 comments published by 84,203 unique users, I capture features of the network topology to understand: whether users participate in various discussions; if discussions fragment in multiple threads showing assortative mixing tendencies; and if comments concentrate around few discussions approaching a power-law like distribution. This paper aims to improve understanding of the argumentative political discourse, but has also applications for political institutions interested in enhancing computer-mediated public reasoning.

KEYWORDS:

• Five Star Movement
• online deliberation
• network analysis
• social media and politics
• political discussion
• political participation
• Italian politics
• Italian general elections 2013
• Beppe Grillo

Notes on contributor

Francesco Bailo is a PhD student in the Department of Government and International Relations at the University of Sydney. He is completing a thesis on how online discussion forums of political organisations are represented as networks; focusing on the forums of the M5S in Italy.

Notes

1. The Wiktionary is a wiki-based open dictionary and – as Wikipedia – a project of the Wikimedia Foundation. I retrieved the gender of each Italian first name present in the dictionary under the category ‘Antroponimi’. See it.wiktionary.org/wiki/Categoria:Antroponimi-IT.

2. Threads are proposals published online by a registered member of the M5S and open to comments.

3. The degree is here intended as the number of connections of each node.

4. Graph generated with the function ba.game of the igraph software package (Csardi and Nepusz 2006)

5. The authors crawled links to fifth depth starting from the most popular videos. That is, the data is biased towards videos that were – or were linked to – the most viewed videos at the time.

6. More precisely a weakly connected component. Anyway since the current network is a bipartite directed network, with links moving only from one to the other type of node, strongly connected components are by definition impossible.

7. Size is here intended as number of nodes.

8. Transformations performed with the functions bipartite.projection and contract.vertices of the igraph software package (Csardi and Nepusz 2006).

9. Algorithm performed with the function fastgreedy.community of the igraph software package (Csardi and Nepusz 2006).

10. In fact, this model is a simple gravity model (with space replaced by time) assuming that the interaction between two populations is directly proportional totheir respective size and inversely proportional to their distance in time (Kolaczyk and Csardi 2014, 162).

11. A 95% confidence interval is the range of values which contains the true unknown value of the parameter – in this case shared_users – with probability .95.

12. The Forum is under the Creative Commons license ‘Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0)’.

13. Disqus is a popular hosting service for blog and Forum comments. It allows to comment over different sites using a unique user identification that can linked to a Disqus account but also to Facebook, Twitter and Google accounts.

14. www.google.com/cse.

15. disqus.com/api/.

16. The results returned by the query were more than 1000 in 72 cases thus requiring the combine use of Google Custom Search Engine and Disqus API for a complete indexing of the Forum.

17. This figure is higher than the total number of threads presented in Section 3 for reasons explained in Section A4 of the appendix.

18. A web crawler should avoid to request pages more quickly than a human would (then not more than 6 pages per minute) to avoid overloading the server and in the worst case scenario making it unable to serve other users.