The Filter Bubble Effect of Algorithmic Ranking on Social Media

Information Systems and Operations Management

Speaker : Guangrui (Kayla) LI

Doctoral Candidate in Information Systems

Hong Kong University of Science and Technology (HKUST)  

HEC Campus - Jouy-En-Josas - Bâtiment V - Salle Bernard Ramanantsoa

 Abstract : 

Although some argue that personalization algorithms can help to deal with information overload and improve user experience, they can also induce algorithmic bias that could skew or limit the information an individual user consumes on the Internet, thus creating the ‘filter bubble’ phenomenon. This study assesses the prevalence of filter bubble on social media, and especially its long-term impact, by making use of a quasi-experiment on two leading social media platforms in China. We leveraged our unique dataset to carry out a difference-in-difference analysis by comparing users’ behavioral changes on Sina and Tencent Weibo. Our findings suggested that after the implementation of ranking algorithms, people’s interest scope became much limited, and this effect was magnified over time. We also found evidence that people encountered less attitude-challenging content after the implementation of ranking algorithms, as shown by their posts becoming less negative in tone. Together, these findings provide strong evidence for the existence of filter bubble caused by ranking algorithm. We further examined the impact of filter bubble on users’ attitude polarization and contribution behavior to the platforms. We found that the filter bubbles increased polarization on platforms by around 15%, and significantly reduced the number of posts generated by users, both reposted and original. Our findings have implications for improved transparency in platform design and for governance of user-generated content.

Bio :  

Guangrui(Kayla) Li is a doctoral candidate in Information Systems, the School of Business and Management at Hong Kong University of Science and Technology (HKUST). Her research interests revolve around the societal impacts of online platform governance strategies (e.g. algorithmic impacts and data piracy) and methodological challenges raised by big data. Her research methodologies include econometrics analyses, statistical analyses, and text mining. Prior to joining HKUST, she received her BSc degree in Peking University.