Negative ties in social networks: the role of contention in group polarization dynamics of the Dutch Black Pete debate

With many of today’s pressing and polarized debates taking place online, specifically on online social networks, there is an unprecedented opportunity to study the dynamics of people self-organizing into groups and agreeing or arguing against each other over important topics. However, previous studies into polarization on social networks have barely considered negative interactions between users: the focus has been on positive interaction, often studied through the lens of networks with positive ties. In this research, we show what the explicit inclusion of negative interactions online can contribute to the study of large scale social dynamics, particularly to that of partisan polarization. We illustrate our approach by a case study of the debate about ’Black Pete’, a caricature from a Dutch tradition with racist connotations, that has divided Dutch society and is an ongoing topic of conflict. We use a dataset of 400.000 Tweets on this topic, many of which contain mentions to specific other users. By means of machine learning, we identify users’ positions in the debate and the sign of user-to-user interactions. Signed network analysis reveals important differences between the group compositions in this network compared to the non signed counterpart, the retweet network, which is typically used for studies into polarization on Twitter. Furthermore, the signs allow for the identification of a rich variety of roles in this debate, defined by combinations ingroup and outgroup sentiments, that stay hidden in an unsigned analysis. Beyond the specific findings of this case study, the present research elucidates the importance of negative interactions in the social realm and illustrates the potential of various research avenues made possible by a signed analysis of contentious debates.