Online Program

BullyBlocker: Integrating Data, Computer, and Psychological Science to Identify Cyberbullying on Social Media (309551)

Keywords: Cyberbullying, Interdisciplinary, Temporal analysis, Hierarchical attention network, Predictive models, Social influence, Personalized models

Increased social media use, particularly among teens, has corresponded with a rise in the prevalence of cyberbullying. Researchers across various disciplines have begun to investigate cyberbullying, yet there has been relatively little work integrating the theoretical frameworks and empirical findings from psychology with the predictive and descriptive models from computer science. The BullyBlocker Project, through the utilization of predictive analytics and analysis of high dimensional data, is oriented around the goal of developing cyberbullying identification models that are informed by psychological science and tested using innovative computer science techniques on large social media datasets. Our specific approaches include incorporating temporal information, personalized user characteristics, social influence components, hierarchical attention networks, and multi-modal data (e.g., images, text, shares) into cyberbullying identification models. Our work underscores the importance of interdisciplinary collaborations that bridge psychological research with advancements in machine learning and the use of big data to identify cyberbullying and reduce its prevalence among teens.