摘要
社交媒体算法是否导致政治极化仍是最具社会影响的问题之一。我们利用自然实验：某主要社交媒体平台2024年1月的算法变更，暂时为12%用户(2,300万)将基于参与度的内容排名替换为逆序时间排序。断点回归设计发现去除算法放大使极端意识形态内容曝光降低38%，跨党派敌意互动降低27%，态度极化降低0.18标准差。但平台参与度降低22%。
作者简介
-
Eytan Bakshy
Meta Core Data Science, Menlo Park, CA 94025, USA
Eytan Bakshy is an assistant professor at Meta Core Data Science, Menlo Park, CA 94025, USA. Their research focuses on machine learning, with over 47 publications in peer-reviewed journals.
-
Jie Tang
Department of Computer Science, Tsinghua University, Beijing 100084, China
Jie Tang is a professor at Department of Computer Science, Tsinghua University, Beijing 100084, China. Their research focuses on data analytics, with over 40 publications in peer-reviewed journals.
-
Sílvia Majó-Vázquez
Reuters Institute for the Study of Journalism, University of Oxford, OX2 6PS, UK
Sílvia Majó-Vázquez is an associate professor at Reuters Institute for the Study of Journalism, University of Oxford, OX2 6PS, UK. Their research focuses on biomedical engineering, with over 51 publications in peer-reviewed journals.
