Dynamics of political Polarization on Twitter: A study of 2017 Turkish Constitutional Referendum

The effects of online social networks on polarization is subject to scholarly attention. Online platforms receive criticism for being catalyzers of echo-chambers, a phenomenon where like-minded individuals form isolated enclaves. Echo-chambers are considered as one of the core drivers of increasing extremism and polarization. Previous research on social networks mostly considers polarization as a state rather than a process. Although a limited number of studies examined the dynamics of polarization, these interpret polarization as a single process that reflects the division between two opposing subgroups. Contrary to the previous work, our study investigates polarization as the outcome of interactions between two separate processes that represent the internal dynamics of two subgroups with conflicting viewpoints. For this purpose, drawing on the phenomenon of echo-chamber, our study examines temporal change in ideological coherence and social cohesion of polarized communities. In Turkey, political polarization is the buzzword that receives attention from academics to politicians. In this context, our study examines the political polarization on Twitter between the supporters of "Yes" and "No" vote, during the period of the 2017 Turkish constitutional referendum. The focus of our attention is the temporal change in opinion and structural dynamics of two opposing communities that represent the sides of the polarization. Although our findings point to a highly segregated network structure between the supporters of "Yes" and "No" vote, significant seasonal behavior differences were also observed between these two subgroups. In general, our findings show that the internal communication network of "Yes" supporters is denser and ideologically more coherent than the internal communication network of "No" supporters during the referendum period, yet a reverse trend is observed in the course of a few days following the referendum date. During this short period of time, the ideological coherence and internal connectivity of "No" supporters were increasing, whereas the ideological coherence and internal connectivity of "Yes" supporters were decreasing and increasing, respectively. In summary, we considered the limitations of macro-level outlook, which views polarization as a single process, in understanding the local dynamics of polarization. Thus, we developed an approach that views polarization as the relative change of two separate processes each of which reflects the internal dynamics of two subgroups in conflict. The results demonstrated the effectiveness of our approach in discovering fine-grained details of polarization.