Fridays 2:30-3:30 pm
One of the major challenges of recommendation systems is the delicate balance between optimizing user preferences and diversity exposure. Failing to balance these priorities results in recommendation systems that push the user population towards a local optimum, a phenomenon that results in “filter bubbles”. We propose a new type of recommendation system that combines evolutionary game theory and cutting edge machine learning to undermine the source of “bubbles”.
Lorenzo Barberis Canonico is a Ph.D. student in human-centered computing at the Clemson University. He is part of the Team Research Analytics in Computational Environments (TRACE) Research Group. His research interest include team cognition, machine learning, cognitive science, and game theory.
Advisor: Dr. Nathan McNeese
Paper: "Bubble Poppers: A Recommendation System to Burst Filter Bubbles," to be submitted to the Nature Machine Intelligence journal mid-November 2019.