Guest Lectures

How Echo Chambers Amplify Cyberbullying: A Cross-Platform Analysis of Social Media Dynamics

Cyberbullying presents a critical and escalating challenge across social media platforms, yet the underlying dynamics of its spread remain insufficiently understood. Emerging evidence suggests that cyberbullying is not merely an individual act but often a collective phenomenon, intensified by group dynamics. In particular, echo chambers defined as tightly-knit online communities where users are primarily exposed to like-minded opinions, reinforcing shared beliefs while excluding dissenting views – create a fertile ground for the amplification of harmful behavior. This study is the first to investigate the role of echo chambers in propagating cyberbullying. Our study utilized a large-scale dataset comprising 2.038 billion comments, linked to 5.691 million posts and short videos, involving interactions from 1.752 billion users across X (formerly Twitter), TikTok, Weibo, and Douyin. Cyberbullying content was identified using a validated cyberbullying detection model and an echo chamber detection algorithm. The results reveal that echo chambers significantly amplify the depth, breadth, speed, and lifecycle (days) of cyberbullying dissemination. Core echo chamber members (high centrality) were key drivers of cyberbullying spread, while peripheral members had limited influence.  Cyberbullying was most prevalent for topics related to society and international affairs and the influence of echo chambers on cyberbullying was found across English and Chinese platforms underscoring echo chambers' role in facilitating harmful behaviors across cultures. These findings position echo chambers as a systematic enabler of online harm offering a foundation for future research and intervention strategies.

Measuring Bias in Human Interactions with Automated Decision-Making Systems and Designing Bias Mitigation Strategies

In this work, Danula will provide an overview of how his research examines user interactions with automated decision-making systems and explores how biases manifest, which influences future system decisions and user perceptions. He will focus on three research streams that he and his team work on. 

• how they use various survey instruments and sensing methods to better understand the user and characterize the interactions, which is instrumental to designing effective interventions.

• how they design to help users focus on their work on multiparty fair recommendations, which addresses the lack of understanding regarding user perceptions of fair recommendations, particularly in contexts where two parties are involved, such as customers and producers in online marketplaces. 

• other exciting work around how they go beyond users and consider the involuntary stakeholders of systems.