The impact of the social media (SM) has been seen on the one hand as the cause of large exacerbation of negative messages, responsible for massively harmful societal phenomenon against democracies. On the other hand, recent studies have begun to look at how these online channels were able to provide a new impulse in human communication. The novelty of our work resides on analysing several axes of polarizations related to different societal topics. We believe our approach to reflect a more complex society, differing from the recent literature, which has considered a unique left-right dichotomic cleavage. Our methodology consists of extracting topics from the priority themes of the SM debate, using BERT language processing techniques and TF-IDF model. Our results show situation of social media interactions in a multidimensional space does exist. We highlight how social media behaviours, polarization and cross-fertilization differ as upon concrete topics. We argue therefore the 'mega-identity partisanship' which differentiate US online users in two different spaces cannot be extended for the rest of countries taking as first evidence the case of Spain. Further research should extend our conclusions for a possible generalization.
Keywords: Artificial intelligence; BERT; Echo-chamber; Group polarization; Multi-cleavage; Social media.
© The Author(s) 2022.