Prosocial Ranking Challenge

July 11, 2024
ozgar_bao.jpg

Congratulations to OSoMe Ph.D. students, Bao Tran Truong and Ozgur Can Seckin, for being named finalists in the Prosocial Ranking Challenge, hosted by the Center for Human Compatible Artificial Intelligence! This competition aimed to develop new social media ranking algorithms that can mitigate potential harms and design systems towards more socially desirable outcomes.

Social media algorithms wield significant influence over what users see, shaping their knowledge, emotions, and interactions. Addressing these impacts, their project pioneered a "bridging" algorithm designed to re-rank social media feeds. The goal is to cultivate civil discourse and appeal to diverse audiences, thereby fostering a more positive and inclusive online environment.

"Our approach focuses on promoting content that sparks diverse and positive engagement, while demoting toxic content and interactions. This is achieved through a dual assessment framework: evaluating posts based on the emotional responses they provoke and the diversity of engagement they attract. Emotional responses are analyzed for polarity and the presence of harmful reactions, while engagement diversity is gauged by the range of ideological perspectives among engaged users," Bao and Ozgur said in a joint statement.

The pair believes this holistic evaluation of posts will help mitigate the negative impacts of recommender algorithms, particularly in terms of affective polarization. This initiative represents a step forward in leveraging AI for socially beneficial ends and underscores OSoMe's commitment to shaping a healthier digital discourse.