The widespread use, virtual nature, and limited regulation of social media enable a range of problematic online behaviors such as cyberbullying that have threatened users' mental health and well-being. Through the integration of computer science and psychology, this project seeks to better understand how problematic online behaviors can be identified and prevented. The research will investigate the connections between specific features of social media and people's well-being. Major components of the work include developing models for the early detection of problematic online behaviors, designing social media features that prioritize users’ well-being, and studying how cyberbullying may elicit bystander intervention. The broader impact of this project includes the creation and sharing of research and educational resources, information for policymakers to raise awareness of problematic online behaviors and effective strategies to address them, and a unique interdisciplinary training experience that helps students develop into scholars who enact meaningful change. The project is structured around three core research aims. The first aim is to study predictive models for early and community-aware detection of problematic online behaviors, with an emphasis on the assessment of severity. It entails the contextualized annotation of social media data and design and testing of machine learning models. The second aim addresses the need for new social media mechanisms that tackle p