User Behavior in Social Media: Engagement, Incivility and Depression

Abstract

User behavior in online social media has been a much researched topic in various fields– and although some aspects of user behavior like political orientation and online harassment have received much of the limelight, some other aspects have remained mostly obscured. In this research we are exploring three specific behavioral aspects: engagement, incivility and mental health; and our ability to predict these aspects. Predicting future engagement of users can be a behavioral research topic, where user-generated contents and activity frequencies can provide valuable insights. These attributes can be used to analyze and predict how civilly users behave in these social platforms, and can also be used to analyze mental health of a user. All three of these behavioral aspects contribute to the health of a community, and have profound influence on the social capital and the sustainability of the social media platforms. We have built prediction models for engagement in multiple social media, and analyzed the features that we have used over a certain period of time and in a cross-platform environment. We built models for identifying incivilities from user-generated contents and used it in social media as uncivil behavior has the potential to effect user engagement in a platform. We built depression detection models from user texts, and introduced a new performance metric that can measure the quality of a prediction model based on its observational latency- and we argue that it is a more expressive metric of an early prediction model than the current state-of-the-art. We believe we have had significant contributions in the fields we have worked on, and have published our works in various conferences and workshops.