BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Memento EPFL// BEGIN:VEVENT SUMMARY:IC Colloquium: Escaping the Echo Chamber: The Quest for Normative News Recommender Systems DTSTART:20211122T161500 DTEND:20211122T171500 DTSTAMP:20260415T024315Z UID:a1d03fd04a816a22dcbf6cdf8b5b784c4f389567d7a168e73c8b3013 CATEGORIES:Conferences - Seminars DESCRIPTION:By: Abraham Bernstein - University of Zurich\nVideo of his tal k\n\nAbstract\nRecommender systems and social networks are often faulted to be the cause for creating Echo Chambers – environments where people mostly encounter news that match their previous choices or those that are popular among similar users\, resulting in their isolation inside fam iliar but insulated information silos. Echo chambers\, in turn\, have bee n attributed to be one cause for the polarization of society\, which lea ds to the increased difficulty to promote tolerance\, build consensus\, a nd forge compromises. To escape these echo chambers\, we propose to chan ge the focus of recommender systems from optimizing prediction accuracy on ly to considering measures for social cohesion.\n\nThis proposition rais es questions in three spheres: In the technical sphere\, we need to inves tigate how to build “socially considerate” recommender systems. To th at end\, we develop a novel recommendation framework with the goal of  improving information diversity using a modified random walk exploration o f the user-item graph. Applying this framework in the context of poli tical content recommendation\, we also propose a new model to estimate the ideological positions for both users and the content they share\, whi ch can recover ideological positions with high accuracy. Based on these  estimated positions\, we generate diversified personalized recommendations .\n\nIn the social sphere\, we need to investigate if the adapted recomme nder systems have the desired effect. To that end\, we present an empiric al pilot study that exposed users to various sets (some diverse) of news .\n\nFinally\, in the normative sphere\, these studies raise the question what kind of diversity is desirable for the functioning of democracy.\n\ nReflecting the consequences of these findings for our discipline\, this talk highlights that computer science needs to increasingly engage with b oth the social and normative challenges of our work\, possibly producing a new understanding of our discipline\n\nBio\nAbraham Bernstein\, Ph.D.\, is a Full Professor of Informatics at the University of Zurich (UZH)\, S witzerland. He received a Diploma in Computer Science from ETH Zurich an d a Ph.D. in Management with a concentration in Information Technologies from the Sloan School of Management at MIT. His research focuses on vario us aspects of the semantic web\, recommender systems\, AI/data mining/ma chine learning\, crowd computing\, and collective intelligence. His work  draws on both technical (computer science\, artificial intelligence)  and social science (organizational psychology/sociology/economics) fou ndations. He is also a founding Director of the University of Zurich's Di gital Society Initiative (DSI) — an interdisciplinary research and teac hing initiative with more than 180 faculty members ranging from divinity to veterinary medicine investigating all aspects of the interplay betwee n society and the digitalization – and President of the Steering Commit tee of the Swiss National Science Foundation's Research Priority Program 77 on the Digital Transformation. He served as a member of the Counci l of Europe's Committee of experts on Human Rights Dimensions of automate d data processing and different forms of artificial intelligence (MSI-A UT) as well as on the editorial boards of various top journals\, includ ing as a co-Editor in Chief at the Journal of Web Semantics or Associate  Editor at the ACM Transactions on Internet Technologies and ACM Transact ions on Interactive Intelligent Systems. He was elected as an individual member of the Swiss Academy of Technical Sciences in 2019.\n\nMore infor mation LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420 https://epfl.zoom.us/ j/61361899823?pwd=K2wvVDBPV2hxVXlKNU9VVU80b0hEUT09 STATUS:CONFIRMED END:VEVENT END:VCALENDAR