Abstract
ABSTRACT Detection of conflict-prone discussions on online social platforms can help moderate debates and prevent negative social processes such as the formation of echo chambers or polarization of opinions. Here, we examined how controversy of a discussion topic can be estimated from formal characteristics of discussion threads on the English Wikipedia. We discovered that dyadic turn-taking patterns tended to convey highly emotional, personal content, disagreement, and words relating to the conflict. Using the fraction of such two-person patterns in the multi-person discussion as a predictor, we were able to classify effective and conflict-prone discussions with 80% accuracy. These results show that monitoring of turn-taking patterns may become one of the easy heuristics that help detect conflicts in task-oriented online groups.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
