An algorithm predicts party switching. From the combination of artificial intelligence and open data a methodology for studying the behaviour of the political class

What are the factors that have driven almost one in three parliamentarians to change parliamentary group in the last 10 years? And is it possible to predict whether an parliamentarian is about to leave his group?
A study published in the journal iScience has developed an algorithm that, by analysing past votes, is able to estimate the probability of an deputy changing parliamentary group. The research is a collaboration between two Institutes of the Scuola Superiore Sant'Anna, the BioRobotics Institute and the Dirpolis Institute, and involved a multidisciplinary team of scholars: prof. Silvestro Micera, the assistant professor Alberto Mazzoni and PhD student Nicolò Meneghetti for the BioRobotics Institute; prof. Emanuele Rossi, prof. Francesca Biondi Dal Monte and affiliate researcher Fabio Pacini for Dirpolis Institute. 

Artificial Intelligence...

The study by Sant'Anna School investigates the relationship between votes expressed within the Italian Chamber of Deputies and the dynamics of party switching in the last two legislative terms (2013-2018 and 2018-2022). The study is based on the combination of two different elements: machine learning algorithms and the possibility to train and test them using voting data from the Chamber of Deputies. This data is publicly available thanks to the Linked Open Data publishing and sharing platform.
These algorithms made it possible with good accuracy to classify and even predict those Deputies who are about to change parliamentary groups as opposed to their colleagues who will remain within their factions. In particular, the algorithm highlighted two elements that predicted the exit from the parliamentary group many weeks in advance: the greater inclination to participate in secret votes than their colleagues, and the concordance level: the Deputy who is about to change parliamentary group shows in fact a progressive decrease in the agreement level with the majority of the group they are about to leave.

… and Open Data

The research was possible thanks to the availability of open data on the institutional website of the Italian Chamber of Deputies. For several years now, on the websites of the Chamber of Deputies and the Senate of the Republic, a great deal of information has been available on the activities of parliamentarians, on bills tabled, amendments, votes and debates. These data are very useful to know more about the activities of our representatives.
This has made it possible to develop an interdisciplinary methodology for analysing votes in the Chamber of Deputies, which has given us a better understanding of certain trends in recent legislatures and in our parliamentary system.

"Although politics has specific criteria and modes of action - explains Emanuele Rossi, Professor of Constitutional Law at the Scuola Sant'Anna - in some circumstances the use of scientific methodologies that are apparently distant from it can help to analyse and predict the behaviour of the political class".

"I am particularly pleased with the results obtained in this study - comments Silvestro Micera, Professor of Electronic Bioengineering at the Scuola Sant'Anna - The combined use of 'open data' and artificial intelligence has and will increasingly have a major impact on the social sciences as well".