| Summary: | Recent studies show that polarised (or, negative emotionally-charged) social media posts lead to the public’s negative attitudes towards democratic institutions and politicians. However, thus far, scholars are yet to definitively grasp how and under what conditions such polarising rhetoric spreads (i.e., it is replicated or augmented) between politicians and the public. This thesis contributes to this gap in the literature by investigating Matteo Salvini’s rhetoric on Twitter, as well as his respondents’ replies. In particular, this research provides a com- putational analysis of the levels of polarised language in both Salvini’s tweets and replies to his tweets, between 2013 and 2019. With a combination of cutting-edge web-scraping methods, machine learning and time-series analyses, I study over 10 million Twitter data- points. First, I find that polarised language spreads bi-directionally from Salvini’s tweets to his respondents, and vice-versa. Second, Salvini’s followers’ replies not only act as an initial trigger to activate polarisation, but also sustain the use of polarised language in both Salvini and respondents across-time. Third, political content matters. Political replies are more likely to be polarised than non-political replies. Moreover, political replies trigger greater polarisation in tweets in the short-term. Instead, political tweets are more likely to display lower levels of polarisation than non-political tweets, and do not have triggering effects on replies’ polarisation in the short-term. These results may shed light on how polarisation affects online behaviour in both politicians and the public. Ultimately, this thesis provides a diverse theoretical and methodological toolkit to further investigate polarisation on Twitter and beyond.
|
|---|