ISE Seminar Series - Marcos M. Vasconcelos

"Recommender Systems that Enrage to Engage"

Today, most information is produced and consumed on digital media platforms, which rely on recommender systems to distribute it. As a result, recommender systems have become a key mechanism for generating massive profits for these platforms. It is widely accepted that most platforms operate by maximizing user engagement, which is then monetized, for example, through advertising. However, engagement is not always aligned with user experience (e.g. enjoyment and well being). In this talk, I will model the tension between engagement and user experience using a simple content and preference disclosure game. I will show that information asymmetry between the recommender system and the user naturally leads to the so-called "enrage to engage" phenomenon: without complete information about the user's private preferences, the recommender system pushes outrageous content. I will also present counterintuitive results showing that a policy under which the platform commits to ignoring users' private preferences can be more profitable (for the platform itself) than requiring users to disclose their preferences when they join. I will close with several open problems and directions for future research. (Joint work with Odilon Camara, USC Marshall School of Business)

Marcos M. Vasconcelos is an Assistant Professor with the Department of Electrical Engineering at the FAMU-FSU College of Engineering, Florida State University. He received his Ph.D. from the University of Maryland, College Park, in 2016. He was a Research Assistant Professor at the Commonwealth Cyber Initiative and the Bradley Department of Electrical and Computer Engineering at Virginia Tech from 2021 to 2022. From 2016 to 2020, he was a Postdoctoral Research Associate in the Ming Hsieh Department of Electrical Engineering at the University of Southern California. His research interests include networked control and estimation, robotic networks, game theory, distributed optimization, distributed learning, and systems biology.