The Polarized Language of Social Media Discussions

Social media's potential to erode democratic foundations through polarization is a growing concern. This talk explores the historical roots of polarization, emphasizing its conversational nature. To better understand this phenomenon, I propose a natural language analysis approach, using machine learning to measure individual linguistic differences and map the polarized structure of social media discussions.

Early findings suggest that language is more sensitive to polarization than social network structures. Power imbalances are reflected in language polarization, with minority viewpoints being more distinctive than majority opinions, and anger rising near the "poles." This research shifts the focus of polarization research from social division to social change, offering new insights into concerns about the potential threats that polarization poses for democracies.

Speaker: Professor Kevin Durrheim, University of Johannesburg

Kevin Durrheim is Distinguished Professor in Psychology at the University of Johannesburg, where he heads the UJ Methods Lab. He is an NRF A-rated scientist and nominated member of the Academy of Science of South Africa. He is a social psychologist with a program of research related to racism, polarization and social change. His co-authored and co-edited books include The Routledge International Handbook of Discrimination, Prejudice and Stereotyping (2021), Qualitative studies of Silence (2019), Race Trouble (2011) and Racial Encounter (2005).