Students: Sasha Rabeno
Project: Understanding the Radicalizing Effects of Recommendation Algorithms | View Poster (PDF)
Major: Integrated Degree in Engineering, Arts & Sciences (IDEAS)
Advisor: Larry Snyder
Abstract
Algorithms that recommend personalized content to users are a staple of nearly all social media platforms. However, these algorithms often push users towards ideologically extreme content, with the potential to escalate from on-screen hate to off-screen violence. This work explores the radicalizing effects of a scoring-based recommendation system, modeled off the deep learning systems in place at YouTube. For a set of users with random video watching preferences (based on a video’s views, length, and extremeness), a score is generated based on weighted parameters to match users with randomly generated videos best aligning with their interests. “Extremeness” is simplified to a quantifiable value between 0 and 1, where 0 and 1 represent opposites of a polarized political extremeness (0.5 would represent a user whose political tastes lie in the center). This scoring function, which uses a video’s length, number of views, extremeness, and proximity to the user’s preferred extremeness as parameters, creates a list of videos for a given user to watch that best align with their preferences. This research finds that with our scoring system in place, users are pushed to watch videos more extreme than they would in the absence of a recommendation system. As more extreme, provoking content increases engagement (and money made) for social media companies, the scoring equation emphasizes a video’s extremeness—which we found to further polarize the videos users watched as this emphasis increases.
About Sasha Rabeno
Sasha Rabeno is a senior in the Integrated Degree in Engineering, Arts & Sciences (IDEAS) honors program from Chester, New Jersey. She has spent the last year as a Clare Boothe Luce Scholar exploring the impacts of large-scale computer algorithms, particularly those developed to solve social problems. Her current research examines how recommendation algorithms, such as YouTube and Facebook, can push users towards ideologically extreme content. This fall, she will be heading to Cornell University to study information science at the graduate level. On-campus, Sasha is a member of the Marching 97, serving as its Manager this past season. Outside of school, Sasha is a competitive Pokémon player and enjoys spending time with her two research assistants (her cats).