Activity: Talk or presentation types › Conference presentation
Description
Manipulative online behaviors such as trolling are becoming increasingly problematic on social media due to their capacity to obfuscate meanings, to influence democratic decision making and to disseminate harmful information. However, trolling is difficult to identify as it involves deception and indirect forms of manipulation. We discuss how analyzing the dynamics of turn-by-turn computer-mediated interactions allows a deeper understanding of trolling in context. We show how Machine Learning (ML) utilizing a theory-based frame for analyzing interaction can achieve good results in trolling identification for text-based asynchronous online conversations.