The value of online mental health communities, how crisis events are described differently over time on social media, and refining how cyberbullying is detected and classified are major topics of research by Georgia Institute of Technology researchers at this week’s International Conference on Web and Social Media (ICWSM), taking place virtually. It was originally scheduled to be held in Atlanta near the Georgia Tech campus.
More than 220 academics at the 14th annual event are convening and discussing work that is especially relevant during a time of an ongoing global health crisis and social unrest that has taken root across the United States.
Research in the conference proceedings includes many topics directly addressing social ills and injustices that are magnified online as well as potential ways to better understand and mitigate them.
Several College of Computing faculty, current and former students, and postdoctoral researchers are part of the organizing committee. Munmun De Choudhury (Interactive Computing) is serving as the general chair of the conference this year. Former Human-Centered Computing Ph.D. student Stevie Chancellor is workshop chair, former Computer Science Ph.D. student Tanushree Mitra is tutorials chair, current CS Ph.D. student Koustuv Saha is web chair, and current postdoc Talayeh Aledavood is local/social chair. GT Computing faculty Diyi Yang (Interactive Computing) and Srijan Kumar (Computational Science and Engineering) are data challenge chairs.
One of the two keynotes at the conference is by IC faculty Amy Bruckman.
Georgia Tech has three papers in this year’s program:
- A study in causal inference by Saha that tests what leads to favorable psychosocial outcomes in mental health forums.
- A paper by human-centered computing Ph.D. student Ian Stewart, with advisors Yang and Jacob Eisenstein, that intends to gather a sharper view of “collective attention” on social media. Looking at descriptive details for a crisis event, researchers find that the information needed to describe that event changes as time goes on.
- A socially-inspired approach to detect cyberbullying online, by incoming Ph.D. student Caleb Ziems. The paper proposes new criteria for cyberbullying (e.g. harmful intent) and finds that both text and social features help prediction. This paper has been recognized with an Honorable Mention Award, given to a total of eight papers this year.
The conference is sponsored by the Association for the Advancement of Artificial Intelligence.